{
  "version": "https://jsonfeed.org/version/1.1",
  "title": "Open-source AI tool releases",
  "home_page_url": "https://github.com/",
  "description": "Events collected by UnlimitedPipe 0.3.2",
  "_unlimitedpipe": {
    "schema": "unlimitedpipe.event/1",
    "generator": "UnlimitedPipe 0.3.2"
  },
  "items": [
    {
      "id": "97f95731a6c871f9797e",
      "title": "langchain-ai/langchain langchain-fireworks==1.6.3",
      "content_text": "Changes since langchain-fireworks==1.6.2\nrelease(fireworks): 1.6.3 (#40834)\nfix(fireworks): declare native PDF inputs unsupported (#40814)\nfix(fireworks): preserve malformed tool arguments as diagnostic JSON (#40818)\nchore(model-profiles): refresh model profile data (#40804)",
      "date_published": "2026-09-25T12:57:38Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "97f95731a6c871f9797e",
          "source": "github",
          "type": "change",
          "key": "https://github.com/langchain-ai/langchain/releases/tag/langchain-fireworks%3D%3D1.6.3",
          "source_url": "https://api.github.com/repos/langchain-ai/langchain/releases",
          "timestamp": null,
          "observed_at": "2026-09-25T17:36:30Z",
          "data": {
            "change": "added",
            "label": "langchain-ai/langchain langchain-fireworks==1.6.3",
            "item_type": "release",
            "summary": "added: langchain-ai/langchain langchain-fireworks==1.6.3",
            "after": {
              "repo": "langchain-ai/langchain",
              "title": "langchain-ai/langchain langchain-fireworks==1.6.3",
              "tag": "langchain-fireworks==1.6.3",
              "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-fireworks%3D%3D1.6.3",
              "published_at": "2026-09-25T12:57:38Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Changes since langchain-fireworks==1.6.2\nrelease(fireworks): 1.6.3 (#40834)\nfix(fireworks): declare native PDF inputs unsupported (#40814)\nfix(fireworks): preserve malformed tool arguments as diagnostic JSON (#40818)\nchore(model-profiles): refresh model profile data (#40804)",
              "assets": 2
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.2"
            },
            {
              "step": "filter",
              "version": "0.3.2",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.2",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.2",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.2",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-fireworks%3D%3D1.6.3"
    },
    {
      "id": "ce754497d07217e7abab",
      "title": "ollama/ollama v0.40.0",
      "content_text": "What's Changed\nModels run on MLX on Apple Silicon by default\nIn this release, model architectures supported by the MLX runner will run by default on Apple Silicon devices.\nollama pull qwen3.8\nollama run qwen3.8\nDuring the RC we will be testing and enabling additional models.\nFull Changelog: https://github.com/ollama/ollama/compare/v0.34.4...v0.40.0-rc0",
      "date_published": "2026-09-25T03:31:52Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "ce754497d07217e7abab",
          "source": "github",
          "type": "change",
          "key": "https://github.com/ollama/ollama/releases/tag/v0.40.0-rc0",
          "source_url": "https://api.github.com/repos/ollama/ollama/releases",
          "timestamp": null,
          "observed_at": "2026-09-25T06:46:01Z",
          "data": {
            "change": "added",
            "label": "ollama/ollama v0.40.0",
            "item_type": "release",
            "summary": "added: ollama/ollama v0.40.0",
            "after": {
              "repo": "ollama/ollama",
              "title": "ollama/ollama v0.40.0",
              "tag": "v0.40.0-rc0",
              "url": "https://github.com/ollama/ollama/releases/tag/v0.40.0-rc0",
              "published_at": "2026-09-25T03:31:52Z",
              "prerelease": true,
              "author": "github-actions[bot]",
              "summary": "What's Changed\nModels run on MLX on Apple Silicon by default\nIn this release, model architectures supported by the MLX runner will run by default on Apple Silicon devices.\nollama pull qwen3.8\nollama run qwen3.8\nDuring the RC we will be testing and enabling additional models.\nFull Changelog: https://github.com/ollama/ollama/compare/v0.34.4...v0.40.0-rc0",
              "assets": 17
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.2"
            },
            {
              "step": "filter",
              "version": "0.3.2",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.2",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.2",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.2",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/ollama/ollama/releases/tag/v0.40.0-rc0"
    },
    {
      "id": "ece746245ac2de0c8be5",
      "title": "langchain-ai/langchain langchain-core==1.6.5",
      "content_text": "Changes since langchain-core==1.6.4\nrelease(core): 1.6.5 (#40816)\nfix(core): abbreviate long tool IDs in XML buffer strings (#40792)",
      "date_published": "2026-09-24T18:11:22Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "ece746245ac2de0c8be5",
          "source": "github",
          "type": "change",
          "key": "https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.6.5",
          "source_url": "https://api.github.com/repos/langchain-ai/langchain/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T19:33:05Z",
          "data": {
            "change": "added",
            "label": "langchain-ai/langchain langchain-core==1.6.5",
            "item_type": "release",
            "summary": "added: langchain-ai/langchain langchain-core==1.6.5",
            "after": {
              "repo": "langchain-ai/langchain",
              "title": "langchain-ai/langchain langchain-core==1.6.5",
              "tag": "langchain-core==1.6.5",
              "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.6.5",
              "published_at": "2026-09-24T18:11:22Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Changes since langchain-core==1.6.4\nrelease(core): 1.6.5 (#40816)\nfix(core): abbreviate long tool IDs in XML buffer strings (#40792)",
              "assets": 2
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.2"
            },
            {
              "step": "filter",
              "version": "0.3.2",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.2",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.2",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.2",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-core%3D%3D1.6.5"
    },
    {
      "id": "145be175f889e8307dcd",
      "title": "langchain-ai/langchain langchain-openai==1.6.6",
      "content_text": "Changes since langchain-openai==1.6.5\nrelease(openai): 1.6.6 (#40800)\nfix(openai): raise on error events in stream path (#40791)",
      "date_published": "2026-09-24T11:31:31Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "145be175f889e8307dcd",
          "source": "github",
          "type": "change",
          "key": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.6.6",
          "source_url": "https://api.github.com/repos/langchain-ai/langchain/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T15:32:58Z",
          "data": {
            "change": "added",
            "label": "langchain-ai/langchain langchain-openai==1.6.6",
            "item_type": "release",
            "summary": "added: langchain-ai/langchain langchain-openai==1.6.6",
            "after": {
              "repo": "langchain-ai/langchain",
              "title": "langchain-ai/langchain langchain-openai==1.6.6",
              "tag": "langchain-openai==1.6.6",
              "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.6.6",
              "published_at": "2026-09-24T11:31:31Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Changes since langchain-openai==1.6.5\nrelease(openai): 1.6.6 (#40800)\nfix(openai): raise on error events in stream path (#40791)",
              "assets": 2
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.6.6"
    },
    {
      "id": "2c7374fdae64e83fde12",
      "title": "langchain-ai/langchain langchain-anthropic==1.7.4",
      "content_text": "Changes since langchain-anthropic==1.7.3\nchore(anthropic): fix integration test cassette (#40790)\nrelease(anthropic): 1.7.4 (#40786)\nfix(anthropic): add Opus 5.5 and GPT-6 profile augmentations (#40785)\nfeat(anthropic,openai): mid-conversation tool changes on SystemMessage (#40758)",
      "date_published": "2026-09-23T17:56:03Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "2c7374fdae64e83fde12",
          "source": "github",
          "type": "change",
          "key": "https://github.com/langchain-ai/langchain/releases/tag/langchain-anthropic%3D%3D1.7.4",
          "source_url": "https://api.github.com/repos/langchain-ai/langchain/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "langchain-ai/langchain langchain-anthropic==1.7.4",
            "item_type": "release",
            "summary": "added: langchain-ai/langchain langchain-anthropic==1.7.4",
            "after": {
              "repo": "langchain-ai/langchain",
              "title": "langchain-ai/langchain langchain-anthropic==1.7.4",
              "tag": "langchain-anthropic==1.7.4",
              "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-anthropic%3D%3D1.7.4",
              "published_at": "2026-09-23T17:56:03Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Changes since langchain-anthropic==1.7.3\nchore(anthropic): fix integration test cassette (#40790)\nrelease(anthropic): 1.7.4 (#40786)\nfix(anthropic): add Opus 5.5 and GPT-6 profile augmentations (#40785)\nfeat(anthropic,openai): mid-conversation tool changes on SystemMessage (#40758)",
              "assets": 2
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-anthropic%3D%3D1.7.4"
    },
    {
      "id": "44a9258d2025f86d7b54",
      "title": "langchain-ai/langchain langchain-openai==1.6.5",
      "content_text": "Changes since langchain-openai==1.6.4\nrelease(openai): 1.6.5 (#40787)\nfix(anthropic): add Opus 5.5 and GPT-6 profile augmentations (#40785)\nfeat(anthropic,openai): mid-conversation tool changes on SystemMessage (#40758)",
      "date_published": "2026-09-23T15:31:52Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "44a9258d2025f86d7b54",
          "source": "github",
          "type": "change",
          "key": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.6.5",
          "source_url": "https://api.github.com/repos/langchain-ai/langchain/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "langchain-ai/langchain langchain-openai==1.6.5",
            "item_type": "release",
            "summary": "added: langchain-ai/langchain langchain-openai==1.6.5",
            "after": {
              "repo": "langchain-ai/langchain",
              "title": "langchain-ai/langchain langchain-openai==1.6.5",
              "tag": "langchain-openai==1.6.5",
              "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.6.5",
              "published_at": "2026-09-23T15:31:52Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Changes since langchain-openai==1.6.4\nrelease(openai): 1.6.5 (#40787)\nfix(anthropic): add Opus 5.5 and GPT-6 profile augmentations (#40785)\nfeat(anthropic,openai): mid-conversation tool changes on SystemMessage (#40758)",
              "assets": 2
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.6.5"
    },
    {
      "id": "044e7d4ff904a170022b",
      "title": "ollama/ollama v0.34.4",
      "content_text": "What's Changed\nStructured outputs on thinking models now apply in a single pass, making them faster and more reliable.\nFixed intermittent \"model not found\" errors with a large local library\nFixed the macOS app becoming unresponsive when checking if ChatGPT or Codex is running.\nQwen 3.8 prompt processing is faster on Apple Silicon.\nGemma 4 on Apple Silicon now picks the best image resolution per image, keeping more detail in high-resolution images.\nUpdated llama.cpp, MLX, and XGrammar.\nFull…",
      "date_published": "2026-09-23T02:24:43Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "044e7d4ff904a170022b",
          "source": "github",
          "type": "change",
          "key": "https://github.com/ollama/ollama/releases/tag/v0.34.4",
          "source_url": "https://api.github.com/repos/ollama/ollama/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "ollama/ollama v0.34.4",
            "item_type": "release",
            "summary": "added: ollama/ollama v0.34.4",
            "after": {
              "repo": "ollama/ollama",
              "title": "ollama/ollama v0.34.4",
              "tag": "v0.34.4",
              "url": "https://github.com/ollama/ollama/releases/tag/v0.34.4",
              "published_at": "2026-09-23T02:24:43Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "What's Changed\nStructured outputs on thinking models now apply in a single pass, making them faster and more reliable.\nFixed intermittent \"model not found\" errors with a large local library\nFixed the macOS app becoming unresponsive when checking if ChatGPT or Codex is running.\nQwen 3.8 prompt processing is faster on Apple Silicon.\nGemma 4 on Apple Silicon now picks the best image resolution per image, keeping more detail in high-resolution images.\nUpdated llama.cpp, MLX, and XGrammar.\nFull Changelog: https://github.com/ollama/ollama/compare/v0.34.3...v0.34.4",
              "assets": 17
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/ollama/ollama/releases/tag/v0.34.4"
    },
    {
      "id": "8ee4b779172f35e4a50e",
      "title": "langchain-ai/langchain langchain-openai==1.6.4",
      "content_text": "Changes since langchain-openai==1.6.3\nrelease(openai): 1.6.4 (#40775)\nchore(model-profiles): refresh openai model profile data (#40774)",
      "date_published": "2026-09-22T22:37:09Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "8ee4b779172f35e4a50e",
          "source": "github",
          "type": "change",
          "key": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.6.4",
          "source_url": "https://api.github.com/repos/langchain-ai/langchain/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "langchain-ai/langchain langchain-openai==1.6.4",
            "item_type": "release",
            "summary": "added: langchain-ai/langchain langchain-openai==1.6.4",
            "after": {
              "repo": "langchain-ai/langchain",
              "title": "langchain-ai/langchain langchain-openai==1.6.4",
              "tag": "langchain-openai==1.6.4",
              "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.6.4",
              "published_at": "2026-09-22T22:37:09Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Changes since langchain-openai==1.6.3\nrelease(openai): 1.6.4 (#40775)\nchore(model-profiles): refresh openai model profile data (#40774)",
              "assets": 2
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openai%3D%3D1.6.4"
    },
    {
      "id": "7e2b40ccb198a9f61d14",
      "title": "vllm-project/vllm v0.30.0",
      "content_text": "v0.30.0\nHighlights\nThis release features 762 commits from 315 contributors (104 new)!\nNew models: DeepSeek-V4.1-Flash (#56214, #56228, #56208) with the whole KV stored in MXFP8 through the FlashMLA V4.1 record on SM100 (#56893), DeepGEMM Mega-mHC (#56962), and async Engram prefetch with Engram DP sharding (#56512); DeepSeek-V4-Flash-Vision-Exp (#54566), also on ROCm (#55107) and with LoRA (#55897); GLM-5.3-Flash (#53906) with EPLB (#55119); K2-Horizon (#55063); Cohere Compass (#54774); Bailing…",
      "date_published": "2026-09-22T05:20:54Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "7e2b40ccb198a9f61d14",
          "source": "github",
          "type": "change",
          "key": "https://github.com/vllm-project/vllm/releases/tag/v0.30.0",
          "source_url": "https://api.github.com/repos/vllm-project/vllm/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "vllm-project/vllm v0.30.0",
            "item_type": "release",
            "summary": "added: vllm-project/vllm v0.30.0",
            "after": {
              "repo": "vllm-project/vllm",
              "title": "vllm-project/vllm v0.30.0",
              "tag": "v0.30.0",
              "url": "https://github.com/vllm-project/vllm/releases/tag/v0.30.0",
              "published_at": "2026-09-22T05:20:54Z",
              "prerelease": false,
              "author": "khluu",
              "summary": "v0.30.0\nHighlights\nThis release features 762 commits from 315 contributors (104 new)!\nNew models: DeepSeek-V4.1-Flash (#56214, #56228, #56208) with the whole KV stored in MXFP8 through the FlashMLA V4.1 record on SM100 (#56893), DeepGEMM Mega-mHC (#56962), and async Engram prefetch with Engram DP sharding (#56512); DeepSeek-V4-Flash-Vision-Exp (#54566), also on ROCm (#55107) and with LoRA (#55897); GLM-5.3-Flash (#53906) with EPLB (#55119); K2-Horizon (#55063); Cohere Compass (#54774); Bailing V3 VL (#55921); Nanbeige4.2 via the Transformers backend (#56071); and a DeepSeek-V4 CPU backend with AVX512/AMX sparse MLA, indexer, mHC and compressor kernels (#55355).\nFast Start: a persistent per-GPU weight-cache daemon holds post-quantized, TP-sharded weights in GPU memory so restarting engines map them over CUDA IPC with --load-format ipccache instead of reloading from disk (#54921), now covering FP4 checkpoints (#55465) and multi-node TP (#55468).\nWatermarking: Gumbel-max watermarked generation and detection with a keyed PRF, per-request opt-out and an example detection endpoint (#54053); dual-key Gumbel-max makes it compatible with speculative decoding (#56122); the Rust frontend forwards the per-request controls (#56338).\nHiSparse: a host-resident tier for sparse-MLA decode that spills KV pages to pinned host memory under GPU pressure and serves top-k misses from a per-request GPU hot buffer, enabled through HiSparseConnector (#53781), with Prometheus counters (#56061), a host cache shared across TP ranks (#56629), and the attention config inferred from the connector (#57041).\nModel Runner V2: dual-batch overlap in eager mode (#50945) and with FULL CUDA graphs for microbatched steps (#51700); MTP (#46994) and EAGLE3/DFlash/DSpark (#50514) speculative decoding under pipeline parallelism; adaptive verification for every draft-model speculator through an online acceptance estimator (#52228); gc frozen during graph capture, cutting capture from 12s to 2s and engine init f",
              "assets": 9
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/vllm-project/vllm/releases/tag/v0.30.0"
    },
    {
      "id": "b92d1498724d1ccb04a9",
      "title": "langchain-ai/langchain langchain-fireworks==1.6.2",
      "content_text": "Changes since langchain-fireworks==1.6.1\nfix(fireworks): use current completions model in LLM tests (#40740)\nhotfix(fireworks): use available model in LLM tests (#40737)\nrelease(fireworks): 1.6.2 (#40735)\nchore(model-profiles): refresh model profile data (#40665)\nchore(deps): bump anyio from 4.11.0 to 4.14.2 in /libs/partners/fireworks (#40639)\nchore(deps): bump urllib3 from 2.7.0 to 2.8.0 in /libs/partners/fireworks (#40587)\nchore(deps): bump langsmith from 0.12.1 to 0.12.6 in…",
      "date_published": "2026-09-22T04:44:04Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "b92d1498724d1ccb04a9",
          "source": "github",
          "type": "change",
          "key": "https://github.com/langchain-ai/langchain/releases/tag/langchain-fireworks%3D%3D1.6.2",
          "source_url": "https://api.github.com/repos/langchain-ai/langchain/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "langchain-ai/langchain langchain-fireworks==1.6.2",
            "item_type": "release",
            "summary": "added: langchain-ai/langchain langchain-fireworks==1.6.2",
            "after": {
              "repo": "langchain-ai/langchain",
              "title": "langchain-ai/langchain langchain-fireworks==1.6.2",
              "tag": "langchain-fireworks==1.6.2",
              "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-fireworks%3D%3D1.6.2",
              "published_at": "2026-09-22T04:44:04Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Changes since langchain-fireworks==1.6.1\nfix(fireworks): use current completions model in LLM tests (#40740)\nhotfix(fireworks): use available model in LLM tests (#40737)\nrelease(fireworks): 1.6.2 (#40735)\nchore(model-profiles): refresh model profile data (#40665)\nchore(deps): bump anyio from 4.11.0 to 4.14.2 in /libs/partners/fireworks (#40639)\nchore(deps): bump urllib3 from 2.7.0 to 2.8.0 in /libs/partners/fireworks (#40587)\nchore(deps): bump langsmith from 0.12.1 to 0.12.6 in /libs/partners/fireworks (#40586)\nchore(deps): bump pygments from 2.20.0 to 2.21.0 in /libs/partners/fireworks (#40585)\nchore(deps): bump idna from 3.19 to 3.20 in /libs/partners/fireworks (#40584)\nchore(model-profiles): refresh model profile data (#40541)\nchore(model-profiles): refresh model profile data (#40500)\nchore(model-profiles): refresh model profile data (#40452)\nchore(model-profiles): refresh model profile data (#40416)\nchore(model-profiles): refresh model profile data (#40399)\nchore(model-profiles): refresh model profile data (#40277)\nchore(model-profiles): refresh model profile data (#40217)\nchore(model-profiles): refresh model profile data (#40198)\nchore(model-profiles): refresh model profile data (#40171)\nchore(model-profiles): refresh model profile data (#40009)\nchore(deps): bump orjson from 3.11.6 to 3.12.0 in /libs/partners/fireworks (#40129)\nchore(deps): bump langsmith from 0.10.16 to 0.12.1 in /libs/partners/fireworks (#40130)",
              "assets": 2
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-fireworks%3D%3D1.6.2"
    },
    {
      "id": "ebf988edd28262e79660",
      "title": "langchain-ai/langchain langchain-openrouter==0.2.9",
      "content_text": "Changes since langchain-openrouter==0.2.8\nrelease(openrouter): 0.2.9 (#40736)\nchore(model-profiles): refresh model profile data (#40705)\nchore(model-profiles): refresh model profile data (#40685)\nchore(model-profiles): refresh model profile data (#40665)\nchore(deps): bump anyio from 4.13.0 to 4.14.2 in /libs/partners/openrouter (#40627)\nchore(model-profiles): refresh model profile data (#40600)\nchore(model-profiles): refresh model profile data (#40541)\nchore(model-profiles): refresh model…",
      "date_published": "2026-09-22T04:16:11Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "ebf988edd28262e79660",
          "source": "github",
          "type": "change",
          "key": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openrouter%3D%3D0.2.9",
          "source_url": "https://api.github.com/repos/langchain-ai/langchain/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "langchain-ai/langchain langchain-openrouter==0.2.9",
            "item_type": "release",
            "summary": "added: langchain-ai/langchain langchain-openrouter==0.2.9",
            "after": {
              "repo": "langchain-ai/langchain",
              "title": "langchain-ai/langchain langchain-openrouter==0.2.9",
              "tag": "langchain-openrouter==0.2.9",
              "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openrouter%3D%3D0.2.9",
              "published_at": "2026-09-22T04:16:11Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Changes since langchain-openrouter==0.2.8\nrelease(openrouter): 0.2.9 (#40736)\nchore(model-profiles): refresh model profile data (#40705)\nchore(model-profiles): refresh model profile data (#40685)\nchore(model-profiles): refresh model profile data (#40665)\nchore(deps): bump anyio from 4.13.0 to 4.14.2 in /libs/partners/openrouter (#40627)\nchore(model-profiles): refresh model profile data (#40600)\nchore(model-profiles): refresh model profile data (#40541)\nchore(model-profiles): refresh model profile data (#40500)\nchore(model-profiles): refresh model profile data (#40452)\nchore(model-profiles): refresh model profile data (#40436)\nchore(model-profiles): refresh model profile data (#40416)\nchore(model-profiles): refresh model profile data (#40399)\nchore(model-profiles): refresh model profile data (#40358)\nchore(model-profiles): refresh model profile data (#40317)\nchore(model-profiles): refresh model profile data (#40277)\nchore(model-profiles): refresh model profile data (#40258)\nchore(model-profiles): refresh model profile data (#40217)\nchore(model-profiles): refresh model profile data (#40198)\nchore(model-profiles): refresh model profile data (#40171)\nchore(model-profiles): refresh model profile data (#40009)\nchore(model-profiles): refresh model profile data (#39954)\nchore(model-profiles): refresh model profile data (#39928)\nchore(model-profiles): refresh model profile data (#39906)\nchore(model-profiles): refresh model profile data (#39875)\nchore(model-profiles): refresh model profile data (#39844)\nchore(model-profiles): refresh model profile data (#39824)\nchore(model-profiles): refresh model profile data (#39789)\nchore(model-profiles): refresh model profile data (#39751)\nchore(model-profiles): refresh model profile data (#39710)\nchore(model-profiles): refresh model profile data (#39692)\nchore(model-profiles): refresh model profile data (#39670)",
              "assets": 2
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/langchain-ai/langchain/releases/tag/langchain-openrouter%3D%3D0.2.9"
    },
    {
      "id": "929b6e69f90d4e78c9fd",
      "title": "open-webui/open-webui v0.11.4",
      "content_text": "Added\n📉 Far smaller slim image. A slim build now comes down at around 175 MB, near enough 89% smaller than the last release, the local models, the packages around them and the tools that installed them all gone from it; what that changes about the way an instance behaves is set out under Changed below and in the documentation. Commit, Commit\n📦 Smaller standard image. The image no longer carries a second copy of Python, two sets of fonts nothing ever loaded, packages nothing imports, or the tool…",
      "date_published": "2026-09-21T19:25:21Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "929b6e69f90d4e78c9fd",
          "source": "github",
          "type": "change",
          "key": "https://github.com/open-webui/open-webui/releases/tag/v0.11.4",
          "source_url": "https://api.github.com/repos/open-webui/open-webui/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "open-webui/open-webui v0.11.4",
            "item_type": "release",
            "summary": "added: open-webui/open-webui v0.11.4",
            "after": {
              "repo": "open-webui/open-webui",
              "title": "open-webui/open-webui v0.11.4",
              "tag": "v0.11.4",
              "url": "https://github.com/open-webui/open-webui/releases/tag/v0.11.4",
              "published_at": "2026-09-21T19:25:21Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Added\n📉 Far smaller slim image. A slim build now comes down at around 175 MB, near enough 89% smaller than the last release, the local models, the packages around them and the tools that installed them all gone from it; what that changes about the way an instance behaves is set out under Changed below and in the documentation. Commit, Commit\n📦 Smaller standard image. The image no longer carries a second copy of Python, two sets of fonts nothing ever loaded, packages nothing imports, or the tool that installed them, taking about 170 MB off a standard build. #29731, #29723, #29725, #29726, #29728, Commit, Commit, Commit, Commit\n🧑‍💻 Skills from a terminal. Skills a connected terminal server offers now sit beside workspace skills everywhere skills are picked — the \"$\" and \"/\" menus, the integrations menu and the skills panel, each marked Terminal — and are used the same way: picking one puts its instructions, its folder and the files it ships with in front of the model, and a model that was only told a skill exists can open it itself. They follow whichever terminal is selected and clear when that changes. Commit\n📔 Terminal instructions file. A model working with a terminal is now handed the AGENTS.md sitting in that terminal's home directory, read afresh at the start of every turn, so the instructions you keep beside your work reach the model without being pasted in. Commit, Commit\n📇 Automatic skill discovery. Every skill you can reach is now listed to the model by name and description, and the full text of one is loaded only when it decides to use it; before, a skill you had not selected in the message box was invisible to it, and this applies to models with built-in tools on. Commit\n🌄 Model background images. A workspace model can now carry a background image, uploaded in its editor and drawn behind the chat whenever that model is selected, sitting below a folder's own background and above your personal one, and it travels with the model through export and import. Com",
              "assets": 0
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/open-webui/open-webui/releases/tag/v0.11.4"
    },
    {
      "id": "5f8f6068d980d8ef1d6a",
      "title": "comfyanonymous/ComfyUI v0.37.0",
      "content_text": "What's Changed\nAimdo 0.5.5 + Auto-detect and enable --fast-disk when the disk is fast (CORE-440) by @rattus128 in https://github.com/Comfy-Org/ComfyUI/pull/16333\n[Partner Nodes] feat(OpenAI): add transparent background support for GPT Image 2 by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/16366\nAdd CFG control to YuE2 Generate Music node. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16373\nAlways put text encoder on GPU when dynamic vram on. by @comfyanonymous in…",
      "date_published": "2026-09-21T07:35:01Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "5f8f6068d980d8ef1d6a",
          "source": "github",
          "type": "change",
          "key": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.37.0",
          "source_url": "https://api.github.com/repos/comfyanonymous/ComfyUI/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "comfyanonymous/ComfyUI v0.37.0",
            "item_type": "release",
            "summary": "added: comfyanonymous/ComfyUI v0.37.0",
            "after": {
              "repo": "comfyanonymous/ComfyUI",
              "title": "comfyanonymous/ComfyUI v0.37.0",
              "tag": "v0.37.0",
              "url": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.37.0",
              "published_at": "2026-09-21T07:35:01Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "What's Changed\nAimdo 0.5.5 + Auto-detect and enable --fast-disk when the disk is fast (CORE-440) by @rattus128 in https://github.com/Comfy-Org/ComfyUI/pull/16333\n[Partner Nodes] feat(OpenAI): add transparent background support for GPT Image 2 by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/16366\nAdd CFG control to YuE2 Generate Music node. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16373\nAlways put text encoder on GPU when dynamic vram on. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16374\n[Partner Nodes] fix(Tripo): refuse a P2 run whose linked GLB or FBX output would be empty by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/16369\nfeat: Support MoGe 3 (CORE-443) by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/16381\nUpdate comfy-kitchen version to 0.2.35 by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16385\nchores: Update default width and height in EmptyLatentImage by @alexisrolland in https://github.com/Comfy-Org/ComfyUI/pull/16384\nchores: Update nodes names and categories by @alexisrolland in https://github.com/Comfy-Org/ComfyUI/pull/16274\nQwen3/3.5/3.8 cudagraphs and w4a8 gemv support (CORE-390) by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/15623\nFix issue with qwen speedup PR. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16389\nLower the pos embed precision of SheetSage2 to match upstream. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16395\n[Partner Nodes] feat(client): send Idempotency-Key on partner-proxy calls and collect replays by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/16220\nBump comfyui-frontend-package to 1.53.6 by @comfy-pr-bot in https://github.com/Comfy-Org/ComfyUI/pull/16386\nFix ace step VAE decode crashing on non bf16 GPUs. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16405\nfeat: Qwen-image 2.1 support (CORE-423) by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/16400\nUpdate embedded d",
              "assets": 4
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.37.0"
    },
    {
      "id": "99eaa7724d4088de4ea4",
      "title": "ollama/ollama v0.34.3",
      "content_text": "What's Changed\nGET /api/show now advertises each model's thinking controls and default:\nAvailable in the CLI with:\nollama show gemma4\nthinking\nlevels false, true\ndefault true\nAvailable in the API with:\nsh\ncurl http://localhost:11434/api/show -d '{\"model\": \"glm-5.3-flash:cloud\"}'\njson\n{\n\"thinking\": {\n\"values\": [\"low\", \"high\", \"max\"],\n\"default\": \"max\"\n}\n}\nAlso available on ollama.com directly for cloud models.\nNemotron H vision models are now supported on Apple Silicon with MLX\nOllama's macOS app…",
      "date_published": "2026-09-19T00:02:57Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "99eaa7724d4088de4ea4",
          "source": "github",
          "type": "change",
          "key": "https://github.com/ollama/ollama/releases/tag/v0.34.3",
          "source_url": "https://api.github.com/repos/ollama/ollama/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "ollama/ollama v0.34.3",
            "item_type": "release",
            "summary": "added: ollama/ollama v0.34.3",
            "after": {
              "repo": "ollama/ollama",
              "title": "ollama/ollama v0.34.3",
              "tag": "v0.34.3",
              "url": "https://github.com/ollama/ollama/releases/tag/v0.34.3",
              "published_at": "2026-09-19T00:02:57Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "What's Changed\nGET /api/show now advertises each model's thinking controls and default:\nAvailable in the CLI with:\nollama show gemma4\nthinking\nlevels false, true\ndefault true\nAvailable in the API with:\nsh\ncurl http://localhost:11434/api/show -d '{\"model\": \"glm-5.3-flash:cloud\"}'\njson\n{\n\"thinking\": {\n\"values\": [\"low\", \"high\", \"max\"],\n\"default\": \"max\"\n}\n}\nAlso available on ollama.com directly for cloud models.\nNemotron H vision models are now supported on Apple Silicon with MLX\nOllama's macOS app will now no longer reopen windows you've closed when activating it\nFix for model pulls from HuggingFace\nFull Changelog: https://github.com/ollama/ollama/compare/v0.34.2...v0.34.3",
              "assets": 17
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/ollama/ollama/releases/tag/v0.34.3"
    },
    {
      "id": "f3054aca84e7115a3b75",
      "title": "comfyanonymous/ComfyUI v0.36.0",
      "content_text": "What's Changed\nAdd new model blueprints and reorganize subgraph categories by @comfyui-wiki in https://github.com/Comfy-Org/ComfyUI/pull/14785\nmain: bump the AMD Windows VA quota to 4TB (CORE-409) by @rattus128 in https://github.com/Comfy-Org/ComfyUI/pull/16199\n[Partner Noes] feat(OpenRouter): add Microsoft mai-image-2.6 models by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/16188\nquantops: drop the dead ROCm triton arch gate by @0xDELUXA in…",
      "date_published": "2026-09-15T22:26:17Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "f3054aca84e7115a3b75",
          "source": "github",
          "type": "change",
          "key": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.36.0",
          "source_url": "https://api.github.com/repos/comfyanonymous/ComfyUI/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "comfyanonymous/ComfyUI v0.36.0",
            "item_type": "release",
            "summary": "added: comfyanonymous/ComfyUI v0.36.0",
            "after": {
              "repo": "comfyanonymous/ComfyUI",
              "title": "comfyanonymous/ComfyUI v0.36.0",
              "tag": "v0.36.0",
              "url": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.36.0",
              "published_at": "2026-09-15T22:26:17Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "What's Changed\nAdd new model blueprints and reorganize subgraph categories by @comfyui-wiki in https://github.com/Comfy-Org/ComfyUI/pull/14785\nmain: bump the AMD Windows VA quota to 4TB (CORE-409) by @rattus128 in https://github.com/Comfy-Org/ComfyUI/pull/16199\n[Partner Noes] feat(OpenRouter): add Microsoft mai-image-2.6 models by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/16188\nquantops: drop the dead ROCm triton arch gate by @0xDELUXA in https://github.com/Comfy-Org/ComfyUI/pull/16211\n[Partner Nodes] feat(Gemini-LLM): Add Gemini 3.8 Flash to the Gemini text node by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/16195\nUpdate instructions for manual install on windows AMD. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16217\n[Partner Nodes] feat(Tripo): migrate to the v3 API, add the Smart Segment node and retire the dead widgets by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/16201\n[Partner Nodes] feat(OpenRouter): add auto aspect ratio to the MAI image node by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/16229\nUpdate workflow templates to v0.11.59 by @comfyui-wiki in https://github.com/Comfy-Org/ComfyUI/pull/16233\nFix h3 fun controlnet with comfy compiler. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16240\nAdd linear to ImageColorSpace. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16247\nImplement Video Concatenate (CORE-436) by @rattus128 in https://github.com/Comfy-Org/ComfyUI/pull/16267\nfix: Fix unit tests for Concatenate Video node by @alexisrolland in https://github.com/Comfy-Org/ComfyUI/pull/16271\nfeat: Marigold v2 support (CORE-431) by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/16232\nSupport Yue2 music model. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/16250\nBump comfyui-frontend-package to 1.52.7 by @christian-byrne in https://github.com/Comfy-Org/ComfyUI/pull/16275\n[Partner Nodes] feat(Bria): add new image edit nodes and the Video Era",
              "assets": 4
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.36.0"
    },
    {
      "id": "0d4ffd546aa20137a71b",
      "title": "ollama/ollama v0.34.2",
      "content_text": "What's Changed\nAdded first-run setup when running ollama, with options to sign in or continue locally. Setup completion is shared with the desktop app on macOS and Windows.\nAdded ollama://apps to open the desktop app’s Apps page directly on macOS and Windows.\nFixed excessive memory growth during long generations with MLX speculative decoding.\nUpdated llama.cpp.\nFull Changelog: https://github.com/ollama/ollama/compare/v0.34.1...v0.34.2",
      "date_published": "2026-09-15T21:21:07Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "0d4ffd546aa20137a71b",
          "source": "github",
          "type": "change",
          "key": "https://github.com/ollama/ollama/releases/tag/v0.34.2",
          "source_url": "https://api.github.com/repos/ollama/ollama/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "ollama/ollama v0.34.2",
            "item_type": "release",
            "summary": "added: ollama/ollama v0.34.2",
            "after": {
              "repo": "ollama/ollama",
              "title": "ollama/ollama v0.34.2",
              "tag": "v0.34.2",
              "url": "https://github.com/ollama/ollama/releases/tag/v0.34.2",
              "published_at": "2026-09-15T21:21:07Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "What's Changed\nAdded first-run setup when running ollama, with options to sign in or continue locally. Setup completion is shared with the desktop app on macOS and Windows.\nAdded ollama://apps to open the desktop app’s Apps page directly on macOS and Windows.\nFixed excessive memory growth during long generations with MLX speculative decoding.\nUpdated llama.cpp.\nFull Changelog: https://github.com/ollama/ollama/compare/v0.34.1...v0.34.2",
              "assets": 17
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/ollama/ollama/releases/tag/v0.34.2"
    },
    {
      "id": "f9e9eb8116f0a4751423",
      "title": "ollama/ollama v0.34.1",
      "content_text": "What's Changed\nMLX safetensors ollama create no longer experimental. GGUF model creation now requires using llama.cpp tooling for safetensor conversion and quantization.\nImproved MLX memory handling on Apple Silicon\nRunaway repeat token detection now requires 100 repeat tokens for reduced false positives (e.g. OCR)\n/api/tags is much faster on large model libraries (3.1 s → 294 ms cold in testing), and model capabilities are now reported consistently.\nDeprecated typicalp: it can no longer be set…",
      "date_published": "2026-09-14T22:14:03Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "f9e9eb8116f0a4751423",
          "source": "github",
          "type": "change",
          "key": "https://github.com/ollama/ollama/releases/tag/v0.34.1",
          "source_url": "https://api.github.com/repos/ollama/ollama/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "ollama/ollama v0.34.1",
            "item_type": "release",
            "summary": "added: ollama/ollama v0.34.1",
            "after": {
              "repo": "ollama/ollama",
              "title": "ollama/ollama v0.34.1",
              "tag": "v0.34.1",
              "url": "https://github.com/ollama/ollama/releases/tag/v0.34.1",
              "published_at": "2026-09-14T22:14:03Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "What's Changed\nMLX safetensors ollama create no longer experimental. GGUF model creation now requires using llama.cpp tooling for safetensor conversion and quantization.\nImproved MLX memory handling on Apple Silicon\nRunaway repeat token detection now requires 100 repeat tokens for reduced false positives (e.g. OCR)\n/api/tags is much faster on large model libraries (3.1 s → 294 ms cold in testing), and model capabilities are now reported consistently.\nDeprecated typicalp: it can no longer be set when creating new models, existing GGUF models retain support.\nMLX and llama.cpp updates\nFull Changelog: https://github.com/ollama/ollama/compare/v0.34.0...v0.34.1-rc1",
              "assets": 17
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/ollama/ollama/releases/tag/v0.34.1"
    },
    {
      "id": "5a3fe6440844cc64d7ee",
      "title": "comfyanonymous/ComfyUI v0.35.0",
      "content_text": "What's Changed\n[Partner Nodes] chore(Google): drop retiring Veo 2 and Veo 3.0 models by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15883\n[Partner Nodes] feat(Recraft): add V4 Styles by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15903\n[Partner Nodes] feat(WAN): add WAN3-Prime model support by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15894\nBump comfyui-frontend-package to 1.51.9 by @comfy-pr-bot in https://github.com/Comfy-Org/ComfyUI/pull/15696\nSupport avif…",
      "date_published": "2026-09-09T19:55:08Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "5a3fe6440844cc64d7ee",
          "source": "github",
          "type": "change",
          "key": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.35.0",
          "source_url": "https://api.github.com/repos/comfyanonymous/ComfyUI/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "comfyanonymous/ComfyUI v0.35.0",
            "item_type": "release",
            "summary": "added: comfyanonymous/ComfyUI v0.35.0",
            "after": {
              "repo": "comfyanonymous/ComfyUI",
              "title": "comfyanonymous/ComfyUI v0.35.0",
              "tag": "v0.35.0",
              "url": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.35.0",
              "published_at": "2026-09-09T19:55:08Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "What's Changed\n[Partner Nodes] chore(Google): drop retiring Veo 2 and Veo 3.0 models by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15883\n[Partner Nodes] feat(Recraft): add V4 Styles by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15903\n[Partner Nodes] feat(WAN): add WAN3-Prime model support by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15894\nBump comfyui-frontend-package to 1.51.9 by @comfy-pr-bot in https://github.com/Comfy-Org/ComfyUI/pull/15696\nSupport avif in Save Image Advanced node. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15891\nfix(video): remux HEVC to mp4/mov as hvc1 via hevcmp4toannexb by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15809\nfix(memory): respect the container cgroup memory limit instead of host RAM (CORE-394) by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15927\nUpdate workflow templates to v0.11.50 by @comfyui-wiki in https://github.com/Comfy-Org/ComfyUI/pull/15931\n[Partner Nodes] feat(Google-Omni): add support for Omni 1.1 model by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15929\nfeat(3d): File3DToMesh node — parse GLB/GLTF/OBJ/STL into MESH by @jtydhr88 in https://github.com/Comfy-Org/ComfyUI/pull/15919\nAdd to readme that we support HDR and high bit depth. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15940\nAdd cfgppud10ab sampler. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15951\nMiniMax-H3: Support PDD LoRA by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/15908\nAdd section about user input tolerance to AGENTS.md by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15955\n[Partner Nodes] fix(HeyGen): update Avatar Video price badge by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15945\nImprove some warning messages. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15977\nfeat: add VideoTrim and VideoCrop nodes with VIDEOEDIT widget inputs by @jtydhr88 in https://github.com/Co",
              "assets": 4
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.35.0"
    },
    {
      "id": "2909b10713ae1d9adccc",
      "title": "huggingface/transformers Release 5.17.0",
      "content_text": "Release v5.17.0\nNew Model additions\nHYV4\nHy4-Preview is a 780B-parameter mixture-of-experts language model that activates 49B parameters per\ntoken. Each MoE layer holds 256 routed experts plus one always-active shared expert and routes every\ntoken to 8 of them. The context window is 1M tokens.\nThe architecture combines four features:\nMulti-head Latent Attention (MLA) compresses keys and values into a low-rank latent\n(kvlorarank) that kvbproj expands back to one key/value per query…",
      "date_published": "2026-09-09T15:42:45Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "2909b10713ae1d9adccc",
          "source": "github",
          "type": "change",
          "key": "https://github.com/huggingface/transformers/releases/tag/v5.17.0",
          "source_url": "https://api.github.com/repos/huggingface/transformers/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "huggingface/transformers Release 5.17.0",
            "item_type": "release",
            "summary": "added: huggingface/transformers Release 5.17.0",
            "after": {
              "repo": "huggingface/transformers",
              "title": "huggingface/transformers Release 5.17.0",
              "tag": "v5.17.0",
              "url": "https://github.com/huggingface/transformers/releases/tag/v5.17.0",
              "published_at": "2026-09-09T15:42:45Z",
              "prerelease": false,
              "author": "vasqu",
              "summary": "Release v5.17.0\nNew Model additions\nHYV4\nHy4-Preview is a 780B-parameter mixture-of-experts language model that activates 49B parameters per\ntoken. Each MoE layer holds 256 routed experts plus one always-active shared expert and routes every\ntoken to 8 of them. The context window is 1M tokens.\nThe architecture combines four features:\nMulti-head Latent Attention (MLA) compresses keys and values into a low-rank latent\n(kvlorarank) that kvbproj expands back to one key/value per query head.\nDeepSeek Sparse Attention (DSA) selects indextopk keys per query with a lightweight indexer.\nFollowing IndexShare, only the layers marked \"full\"\nin indexertypes run an indexer; \"shared\" layers reuse the previous full layer's selection.\nGated MLA with learnable attention sinks, where each head owns a sink logit that participates\nin the softmax and contributes no value, as in GPT-OSS.\nIndependent Hyper-Connections (iHC) replace the plain residual path with hcmult parallel\nresidual streams that are collapsed before, and redistributed after, every sublayer.\nThe implementation does not execute the multi-token prediction (MTP) layers. Released checkpoints\nkeep those weights so that other runtimes can use them for speculative decoding; they are ignored\nat load time.\nLinks: Documentation\nAdd h4 (#48473) by @ArthurZucker in #48473\nVibeVoice\nVibeVoice is a novel framework for synthesizing high-fidelity, long-form speech with multiple speakers by employing a next-token diffusion approach within a Large Language Model (LLM) structure. It's designed to capture the authentic conversational \"vibe\" and is particularly suited for generating audio content like podcasts and multi-participant audiobooks.\nLinks: Documentation\nImplement VibeVoice (#40546) by @pengzhiliang in #40546\nNeoMME\nNeoMME is a family of efficient 260M and 800M parameter multimodal-native multilingual foundation encoders from H Company. It processes multilingual text tokens and raw image patches in a single bidirectional Transformer",
              "assets": 0
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/huggingface/transformers/releases/tag/v5.17.0"
    },
    {
      "id": "5575d3b9e1a7e79e3f4b",
      "title": "vllm-project/vllm v0.29.0",
      "content_text": "v0.29.0\nHighlights\nThis release features 594 commits from 277 contributors (91 new)!\nModel Runner V2 is now the default for all models (#53183), completing the rollout that began with pooling models (#48290). MRV2 also gained CUDA graph memory profiling for KV cache auto-sizing (#53306), batch-sharded sampling that cuts per-step logits memory by 1/TP (#50465), prompt embeds (#42963), extracthiddenstates speculation (#49811), padded FULL cudagraph dispatch for uniform decode under spec decode…",
      "date_published": "2026-09-09T08:54:49Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "5575d3b9e1a7e79e3f4b",
          "source": "github",
          "type": "change",
          "key": "https://github.com/vllm-project/vllm/releases/tag/v0.29.0",
          "source_url": "https://api.github.com/repos/vllm-project/vllm/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "vllm-project/vllm v0.29.0",
            "item_type": "release",
            "summary": "added: vllm-project/vllm v0.29.0",
            "after": {
              "repo": "vllm-project/vllm",
              "title": "vllm-project/vllm v0.29.0",
              "tag": "v0.29.0",
              "url": "https://github.com/vllm-project/vllm/releases/tag/v0.29.0",
              "published_at": "2026-09-09T08:54:49Z",
              "prerelease": false,
              "author": "khluu",
              "summary": "v0.29.0\nHighlights\nThis release features 594 commits from 277 contributors (91 new)!\nModel Runner V2 is now the default for all models (#53183), completing the rollout that began with pooling models (#48290). MRV2 also gained CUDA graph memory profiling for KV cache auto-sizing (#53306), batch-sharded sampling that cuts per-step logits memory by 1/TP (#50465), prompt embeds (#42963), extracthiddenstates speculation (#49811), padded FULL cudagraph dispatch for uniform decode under spec decode (#53407), and DP-sync skipping before EAGLE/MTP draft prefill (#53694). MRV1 remains in use for a few ROCm models and features MRV2 does not yet support.\nNew models: Hy4-preview, Tencent's 770B/49B-active MoE with Gated DeepSeek Sparse Attention and native MTP (#54160); Qwen3.8-Flash-Next with BF16/FP8/NVFP4 and MTP (#53896); GraniteSWA and GraniteMoeSWA (#52706); NemotronHOmniReasoningV3 with MTP (#52929, #53121); Kimi K3 NVFP4 checkpoints (#53132).\nKimi-K3 and DeepSeek V4 performance: fused MXFP4 top-k finalization in the K3 latent tail (about 5% E2E latency, #53152), K3 Mamba metadata preparation in one Triton launch (6.6-7.6x kernel speedup, #52388), tuned Hopper low-latency GEMM (#54088) now also dispatched on SM100 (#53534) and used for ehproj (12.9-25.2% kernel speedup, #53942), GEMM-RS extended to GEMM-AR (#53053), MLA gate merged into the QKV-A projection (#54015), K3 DCP with DSpark (#52188) and DCP partial prefix cache hits (#50493); DeepSeek V4 shared experts fused into MegaMoE (#53040), adaptive top-k width re-landed (#52823), a native SwiGLU clamp kernel for Humming MoE (#53685), and an opt-in FlashInfer moeep expert backend (#49636).\nSpeculative decoding: per-request acceptance stats in OpenAI API responses via --per-request-spec-decode-metrics (#48915), adaptive verification extended to logprobs (#52242), SM100 sparse MLA for GLM-5.2 (#52783) and DeepSeek V4 on SM90 (#52795), Qwen3-Omni DSpark drafts (#52560), PLaMo3 EAGLE-3/DFlash (#54239), and DFlash2 loading f",
              "assets": 9
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/vllm-project/vllm/releases/tag/v0.29.0"
    },
    {
      "id": "0c9dbfacce9bf358675d",
      "title": "ollama/ollama v0.34.0",
      "content_text": "Use Ollama models in ChatGPT Desktop\nOllama models can now be used directly in ChatGPT Desktop, so you can keep your existing workflow while running open models. Setup is available from the Ollama app on MacOS.\nThis release also improves structured output performance on Apple Silicon, adds support for OpenAI-compatible client tool search and response compaction.\nFull Changelog: https://github.com/ollama/ollama/compare/v0.33.3...v0.34.0",
      "date_published": "2026-09-05T23:49:00Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "0c9dbfacce9bf358675d",
          "source": "github",
          "type": "change",
          "key": "https://github.com/ollama/ollama/releases/tag/v0.34.0",
          "source_url": "https://api.github.com/repos/ollama/ollama/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "ollama/ollama v0.34.0",
            "item_type": "release",
            "summary": "added: ollama/ollama v0.34.0",
            "after": {
              "repo": "ollama/ollama",
              "title": "ollama/ollama v0.34.0",
              "tag": "v0.34.0",
              "url": "https://github.com/ollama/ollama/releases/tag/v0.34.0",
              "published_at": "2026-09-05T23:49:00Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Use Ollama models in ChatGPT Desktop\nOllama models can now be used directly in ChatGPT Desktop, so you can keep your existing workflow while running open models. Setup is available from the Ollama app on MacOS.\nThis release also improves structured output performance on Apple Silicon, adds support for OpenAI-compatible client tool search and response compaction.\nFull Changelog: https://github.com/ollama/ollama/compare/v0.33.3...v0.34.0",
              "assets": 17
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/ollama/ollama/releases/tag/v0.34.0"
    },
    {
      "id": "a6725ed4adb394530e8a",
      "title": "open-webui/open-webui v0.11.3",
      "content_text": "Added\n♿ Accessibility mode reaches the menus. Accessibility mode now marks the menu entry you are pointing at and the model already chosen with a stronger background, across the dropdown menus, their submenus, and the model picker together with its filter and compare controls, so those cues carry the contrast the accessibility guidelines ask for in both themes. Commit, Commit\n🔄 General improvements. Various improvements were implemented across the application to enhance performance, stability…",
      "date_published": "2026-08-31T14:55:53Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "a6725ed4adb394530e8a",
          "source": "github",
          "type": "change",
          "key": "https://github.com/open-webui/open-webui/releases/tag/v0.11.3",
          "source_url": "https://api.github.com/repos/open-webui/open-webui/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "open-webui/open-webui v0.11.3",
            "item_type": "release",
            "summary": "added: open-webui/open-webui v0.11.3",
            "after": {
              "repo": "open-webui/open-webui",
              "title": "open-webui/open-webui v0.11.3",
              "tag": "v0.11.3",
              "url": "https://github.com/open-webui/open-webui/releases/tag/v0.11.3",
              "published_at": "2026-08-31T14:55:53Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Added\n♿ Accessibility mode reaches the menus. Accessibility mode now marks the menu entry you are pointing at and the model already chosen with a stronger background, across the dropdown menus, their submenus, and the model picker together with its filter and compare controls, so those cues carry the contrast the accessibility guidelines ask for in both themes. Commit, Commit\n🔄 General improvements. Various improvements were implemented across the application to enhance performance, stability, and security.\n🌐 Translation updates. Translations for Indonesian were enhanced and expanded.\nFixed\n💥 Chat branches stay connected after reloads. A reply saved under an earlier message now stays listed under that message, so branch arrows, exports, reloads, and later edits keep the whole conversation in view, and chats already saved with that link missing are repaired when opened. #29299\n🧱 Upgrades fail clearly instead of starting half updated. A failed database upgrade now stops at the migration error that caused it, instead of starting anyway and reporting a missing table or column such as 'chat.timerat' later, which is the upgrade failure seen after moving from 0.11.0, 0.11.1, or 0.11.2. #29280\n🔤 Custom interface fonts reach more of the app. The font chosen in interface settings now applies to dropdowns and other interface text that previously fell back to the standard font. Commit\n🔌 Disconnect OAuth only where there is OAuth. The disconnect control on a tool server reached over MCP now appears only where that server signs in through OAuth and an account is connected, rather than on servers that use no sign-in at all. #29296",
              "assets": 0
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/open-webui/open-webui/releases/tag/v0.11.3"
    },
    {
      "id": "5163c9e26e5c10090e04",
      "title": "open-webui/open-webui v0.11.2",
      "content_text": "Added\n🖼️ Richer previews for terminal files. Word documents and slide decks produced in the terminal are now previewed as the finished document rather than an approximation, and every document preview gains a page strip down the side with numbered thumbnails you can click to jump straight to a page, and the notice warning that a preview might differ from the download is gone now that it does not. Commit, Commit, Commit, Commit, Commit\n⚡ Less overhead on every message. Deployments without…",
      "date_published": "2026-08-31T05:48:23Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "5163c9e26e5c10090e04",
          "source": "github",
          "type": "change",
          "key": "https://github.com/open-webui/open-webui/releases/tag/v0.11.2",
          "source_url": "https://api.github.com/repos/open-webui/open-webui/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "open-webui/open-webui v0.11.2",
            "item_type": "release",
            "summary": "added: open-webui/open-webui v0.11.2",
            "after": {
              "repo": "open-webui/open-webui",
              "title": "open-webui/open-webui v0.11.2",
              "tag": "v0.11.2",
              "url": "https://github.com/open-webui/open-webui/releases/tag/v0.11.2",
              "published_at": "2026-08-31T05:48:23Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Added\n🖼️ Richer previews for terminal files. Word documents and slide decks produced in the terminal are now previewed as the finished document rather than an approximation, and every document preview gains a page strip down the side with numbered thumbnails you can click to jump straight to a page, and the notice warning that a preview might differ from the download is gone now that it does not. Commit, Commit, Commit, Commit, Commit\n⚡ Less overhead on every message. Deployments without pipelines configured, which is the default, no longer pay setup work for them on each chat message and background task. #29146\n🏎️ Lighter model list refreshes. On deployments running several workers, a refresh that finds the model list unchanged no longer has every worker rewrite the whole list to the shared cache, cutting the work and the traffic each refresh costs. #29264\n🧵 Smoother busy websocket servers. Large Redis-backed websocket deployments now spend far less time sweeping old sessions, reading their own session data back from Redis, and checking every outgoing websocket message for file attachments Open WebUI does not send, so channel posts, collaboration updates, heartbeats, and live chat updates put less load on busy servers. #28835, #28180\n🧰 A new request filter step. Filter authors can now use request to adjust the payload right before each model call, including follow-up calls after tool use, while existing inlet, stream, and outlet filters keep working as before. Commit, Commit\n🤏 More room in file previews on touch screens. File previews on phones and tablets no longer show zoom buttons that sit over an already small preview, leaving pinch to zoom to do the job. #29176, #29152\n👆 Resizing panels by touch. The divider beside the main sidebar or a side panel can now be dragged on a touchscreen or with a stylus, and a drag made with the mouse keeps following the pointer when it leaves the window. Commit, Commit, Commit\n🔤 Choice of interface font. Interface settings now of",
              "assets": 0
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/open-webui/open-webui/releases/tag/v0.11.2"
    },
    {
      "id": "483828ef42f6ee623b98",
      "title": "huggingface/transformers Release v5.16.1",
      "content_text": "Release v5.16.1\nThis is a special release as we include GLM! (and a few small fixes)\nGLM-5.3-Flash\nGLM-5.3-Flash, the first natively multimodal model in the GLM-5 series. With 320B total parameters and just 18B active parameters, it outperforms GLM-5.2 across benchmarks and real-world workloads at one-tenth the price, while approaching Claude Opus 4.8 on coding and agentic benchmarks.\nGLM-5.3-Flash starts from a newly trained base model, with its architecture and training recipe redesigned…",
      "date_published": "2026-08-26T14:50:01Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "483828ef42f6ee623b98",
          "source": "github",
          "type": "change",
          "key": "https://github.com/huggingface/transformers/releases/tag/v5.16.1",
          "source_url": "https://api.github.com/repos/huggingface/transformers/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "huggingface/transformers Release v5.16.1",
            "item_type": "release",
            "summary": "added: huggingface/transformers Release v5.16.1",
            "after": {
              "repo": "huggingface/transformers",
              "title": "huggingface/transformers Release v5.16.1",
              "tag": "v5.16.1",
              "url": "https://github.com/huggingface/transformers/releases/tag/v5.16.1",
              "published_at": "2026-08-26T14:50:01Z",
              "prerelease": false,
              "author": "vasqu",
              "summary": "Release v5.16.1\nThis is a special release as we include GLM! (and a few small fixes)\nGLM-5.3-Flash\nGLM-5.3-Flash, the first natively multimodal model in the GLM-5 series. With 320B total parameters and just 18B active parameters, it outperforms GLM-5.2 across benchmarks and real-world workloads at one-tenth the price, while approaching Claude Opus 4.8 on coding and agentic benchmarks.\nGLM-5.3-Flash starts from a newly trained base model, with its architecture and training recipe redesigned around capability and efficiency. For the first time in the GLM series, we introduce a hybrid architecture combining sparse and linear attention, sharply reducing long-context serving costs while preserving precise long-context capabilities. The model also adopts Manifold-Constrained Hyper-Connections (mHC) to further improve scaling efficiency. Together with our latest 30T-token multimodal pre-training corpus, these changes enable GLM-5.3-Flash to deliver more intelligence with less compute.\nLinks: Documentation\n[Glm 5.3 Flash] GLM 5.3 Flash Support (#48342) by @Dovis01 in #48342\nSmall patch fixes\nMainly BC behavior for TP and pinning a hf kernel for security reasons :hugs:\nRestore BC for the tensor-parallel API (#48300) by @ArthurZucker\nFix kernel commit and repo paths for ESMFold2 (#48186) by @Rocketknight1\nFull Changelog: https://github.com/huggingface/transformers/compare/v5.16.0...v5.16.1",
              "assets": 0
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/huggingface/transformers/releases/tag/v5.16.1"
    },
    {
      "id": "6769447ad40052077fed",
      "title": "huggingface/transformers Release: v5.16.0",
      "content_text": "Release v5.16.0\nNew Model additions\nQwen4-Exp\nQwen4-Exp builds on Qwen3.5's hybrid text and multimodal architecture with three key components: GatedResidual (GR), Qwen Sparse Attention (QSA), and Per-Layer Embedding (PLE).\nGR is a Qwen-developed residual architecture that combines Hyper-Connection with GatedNorm. It mixes multiple residual streams with fine-grained elementwise gating before each attention and Mixture-of-Experts (MoE) block, then controls how much of the block output is injected…",
      "date_published": "2026-08-26T12:35:15Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "6769447ad40052077fed",
          "source": "github",
          "type": "change",
          "key": "https://github.com/huggingface/transformers/releases/tag/v5.16.0",
          "source_url": "https://api.github.com/repos/huggingface/transformers/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "huggingface/transformers Release: v5.16.0",
            "item_type": "release",
            "summary": "added: huggingface/transformers Release: v5.16.0",
            "after": {
              "repo": "huggingface/transformers",
              "title": "huggingface/transformers Release: v5.16.0",
              "tag": "v5.16.0",
              "url": "https://github.com/huggingface/transformers/releases/tag/v5.16.0",
              "published_at": "2026-08-26T12:35:15Z",
              "prerelease": false,
              "author": "Cyrilvallez",
              "summary": "Release v5.16.0\nNew Model additions\nQwen4-Exp\nQwen4-Exp builds on Qwen3.5's hybrid text and multimodal architecture with three key components: GatedResidual (GR), Qwen Sparse Attention (QSA), and Per-Layer Embedding (PLE).\nGR is a Qwen-developed residual architecture that combines Hyper-Connection with GatedNorm. It mixes multiple residual streams with fine-grained elementwise gating before each attention and Mixture-of-Experts (MoE) block, then controls how much of the block output is injected back into each stream.\nQSA uses multiple query heads to score compressed key blocks, selects the most relevant contiguous token blocks, and keeps the incomplete trailing block uncompressed. This block-level selection reduces indexing overhead and improves memory locality for long sequences. Combined with Gated DeltaNet, QSA makes Qwen4-Exp the first hybrid architecture to integrate linear and sparse attention, substantially improving inference efficiency for long-context workloads.\nPLE enriches selected decoder layers with layer-specific lexical features derived from hashed token n-grams and a dilated depthwise convolution.\nLinks: Documentation\nAdd Qwen4Exp model (#48337) by @Cyrilvallez in #48337\nGraniteSpeech5\nGranite Speech 5.0 Turbo CTC is a lightweight (470M parameters) conformer encoder for automatic speech recognition, trained with Connectionist Temporal Classification (CTC) on BPE targets. It is a fast, encoder-only member of the Granite Speech family: transcription requires a single forward pass followed by greedy CTC decoding, with no autoregressive decoder.\nArchitecturally, it extends the Granite Speech conformer CTC encoder with:\nFrame stacking + block-wise time subsampling: the feature extractor stacks pairs of log-mel(+delta) frames (2x), and the first two conformer blocks each subsample time by 2 through a stride-2 depthwise convolution (with a mean-pooled residual), for a total 8x time reduction at 10 ms mel hop.\nBlock attention with Shaw's relative positional",
              "assets": 0
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/huggingface/transformers/releases/tag/v5.16.0"
    },
    {
      "id": "f76be6f5d4aa0827db82",
      "title": "vllm-project/vllm v0.28.0",
      "content_text": "v0.28.0\nHighlights\nThis release features 584 commits from 270 contributors (76 new)!\nKimi-K3 performance push: a major optimization effort for Kimi-K3 across the stack — Decode Context Parallel (DCP) support (#50484), fused FlashKDA decode and prefill kernels (#50654, #51311, #52458), SiTU activation support for MegaMoE (#50510), GEMM-RS for sequence parallelism (#52079), combined all-gathers with 1.53x kernel-level speedup (#51070), an adaptive speculative token budget delivering 60% better…",
      "date_published": "2026-08-26T09:46:30Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "f76be6f5d4aa0827db82",
          "source": "github",
          "type": "change",
          "key": "https://github.com/vllm-project/vllm/releases/tag/v0.28.0",
          "source_url": "https://api.github.com/repos/vllm-project/vllm/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "vllm-project/vllm v0.28.0",
            "item_type": "release",
            "summary": "added: vllm-project/vllm v0.28.0",
            "after": {
              "repo": "vllm-project/vllm",
              "title": "vllm-project/vllm v0.28.0",
              "tag": "v0.28.0",
              "url": "https://github.com/vllm-project/vllm/releases/tag/v0.28.0",
              "published_at": "2026-08-26T09:46:30Z",
              "prerelease": false,
              "author": "khluu",
              "summary": "v0.28.0\nHighlights\nThis release features 584 commits from 270 contributors (76 new)!\nKimi-K3 performance push: a major optimization effort for Kimi-K3 across the stack — Decode Context Parallel (DCP) support (#50484), fused FlashKDA decode and prefill kernels (#50654, #51311, #52458), SiTU activation support for MegaMoE (#50510), GEMM-RS for sequence parallelism (#52079), combined all-gathers with 1.53x kernel-level speedup (#51070), an adaptive speculative token budget delivering 60% better DSpark TTFT (#51725), and optional shared-expert sharding saving 17 GiB of memory per GPU (#50912). Kimi-K3 also now runs on ROCm with the V2 model runner (#51653).\nDeepSeek V4: sparse MLA now works end-to-end for plain decode, MTP, and DSpark speculative decoding (#51538), joined by AMD Quark NVFP4 support (#47972), reasoning-effort prompts and mappings (#50580), sparse top-k metadata kernel optimizations (#52084, #51967), narrowed eager CUDA graph regions (#51430, #52401), and ROCm enablement on gfx11 and gfx950 (#47017, #52212).\nSpeculative decoding advances: DFlash2 with local convolution and a candidate selector (#52816), DSpark confidence-scheduled verification (#47808), and async scheduling auto-enabled for draft models (#48341).\nModel Runner V2 maturation: E/P/D disaggregation (#38390), weight offloading (#51413), multi-layer MTP KV cache support (#50062), encoder CUDA graphs (#49852), decoder token-wise pooling (#50931) plus Transformers pooling models (#52425), attention-free models (#52374), and thinkingtokenbudget support (#46727).\nTiered KV cache offloading: disk offloading support (#49644), out-of-tree secondary tier managers via modulepath (#51007), partial secondary-tier load results (#50321), tiering metrics (#48798), and a canonical CPU layout for parallelism-agnostic offload (#48414).\nRust frontend & gRPC: a standalone renderer (#50289), multimodal image inference over gRPC (#50368), explicit data-parallel rank routing (#51178), and RL lifecycle control (#5131",
              "assets": 9
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/vllm-project/vllm/releases/tag/v0.28.0"
    },
    {
      "id": "5d0b316b0d55065d7636",
      "title": "comfyanonymous/ComfyUI v0.34.0",
      "content_text": "What's Changed\nFix minimax music not working on non dynamic vram. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15588\nAdd MiniMaxH3AddGuide for anchoring image and audio guides at any frame by @drozbay in https://github.com/Comfy-Org/ComfyUI/pull/15439\nchore(openapi): sync shared API contract from cloud@94d0f1b by @comfy-pr-bot in https://github.com/Comfy-Org/ComfyUI/pull/15041\nBump comfyui-frontend-package to 1.49.6 by @comfy-pr-bot in…",
      "date_published": "2026-08-26T02:09:08Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "5d0b316b0d55065d7636",
          "source": "github",
          "type": "change",
          "key": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.34.0",
          "source_url": "https://api.github.com/repos/comfyanonymous/ComfyUI/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "comfyanonymous/ComfyUI v0.34.0",
            "item_type": "release",
            "summary": "added: comfyanonymous/ComfyUI v0.34.0",
            "after": {
              "repo": "comfyanonymous/ComfyUI",
              "title": "comfyanonymous/ComfyUI v0.34.0",
              "tag": "v0.34.0",
              "url": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.34.0",
              "published_at": "2026-08-26T02:09:08Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "What's Changed\nFix minimax music not working on non dynamic vram. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15588\nAdd MiniMaxH3AddGuide for anchoring image and audio guides at any frame by @drozbay in https://github.com/Comfy-Org/ComfyUI/pull/15439\nchore(openapi): sync shared API contract from cloud@94d0f1b by @comfy-pr-bot in https://github.com/Comfy-Org/ComfyUI/pull/15041\nBump comfyui-frontend-package to 1.49.6 by @comfy-pr-bot in https://github.com/Comfy-Org/ComfyUI/pull/15526\nSpeedup Gemma4 text generation (CORE-371) by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/15054\nUpdate embedded docs to v0.5.10 by @comfyui-wiki in https://github.com/Comfy-Org/ComfyUI/pull/15613\nFix thinking handling by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/15611\nEnable dynamic vram by default on ROCm 7.14 and higher. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15633\n[Partner Nodes] feat(ByteDance): add Seedance 2.5 tasktype for video extension by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15579\n[Partner Nodes] Stop adding an opaque alpha channel to API node images by @christian-byrne in https://github.com/Comfy-Org/ComfyUI/pull/15369\nfix(tests): accept Python 3.14 math error messages by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15645\nAdd Minimax Music 3 to readme. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15656\n[Partner Nodes] feat(x-comfy-credits): remove custom price extractor logic by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15655\nAllow regular single image Empty Latent Image node to be used with H3. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15677\n[Partner Nodes] chore(Kling): remove kling-v2 image model by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15676\nForward node class attributes into schema for dataset nodes by @christian-byrne in https://github.com/Comfy-Org/ComfyUI/pull/15683\n[Partner Nodes] feat(FishAudio): implement basi",
              "assets": 4
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.34.0"
    },
    {
      "id": "978ce3fb151dbfcba8df",
      "title": "open-webui/open-webui v0.11.1",
      "content_text": "Added\n🚦 Human in the loop tool approval. Where an administrator has turned it on, you can switch a conversation from letting tools run freely to being asked first, so a model that wants to use a tool stops and waits for you to allow or deny it, one call at a time in a saved conversation, by button or by keyboard shortcut, with your choice remembered for this conversation and for future ones, switching back to running freely releasing anything already waiting, and automations, channel replies…",
      "date_published": "2026-08-25T21:17:57Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "978ce3fb151dbfcba8df",
          "source": "github",
          "type": "change",
          "key": "https://github.com/open-webui/open-webui/releases/tag/v0.11.1",
          "source_url": "https://api.github.com/repos/open-webui/open-webui/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "open-webui/open-webui v0.11.1",
            "item_type": "release",
            "summary": "added: open-webui/open-webui v0.11.1",
            "after": {
              "repo": "open-webui/open-webui",
              "title": "open-webui/open-webui v0.11.1",
              "tag": "v0.11.1",
              "url": "https://github.com/open-webui/open-webui/releases/tag/v0.11.1",
              "published_at": "2026-08-25T21:17:57Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Added\n🚦 Human in the loop tool approval. Where an administrator has turned it on, you can switch a conversation from letting tools run freely to being asked first, so a model that wants to use a tool stops and waits for you to allow or deny it, one call at a time in a saved conversation, by button or by keyboard shortcut, with your choice remembered for this conversation and for future ones, switching back to running freely releasing anything already waiting, and automations, channel replies, and temporary chats unaffected. Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit\n🙋‍♂️ Models that can ask you a question. A new built-in tool lets a model pause and put up to three multiple-choice questions to you before continuing, with room to type your own answer instead, and the question survives a reload in a saved conversation, so you can come back and answer it later rather than losing the conversation. Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit\n🖇 Agents can now display terminal files directly. A model can now show a file it made in a terminal directly in its reply, with a preview and a download button, instead of describing a path that led nowhere when clicked, and a new interface setting chooses whether these open in the reply or in the files pane. Commit, Commit, Commit, #27650\n📶 Streaming rebuilt from the ground up. A reply now streams as small pieces of new text instead of resending the whole message so far with every update, so the data sent over a reply grows with its length rather than with its length squared, which on a server with many people chatting at once means far less processor time spent encoding, passing, and decoding those updates, far less load and memory on the shared cache that carries them between instances, and far less work in your browser, which no longer takes in the whole reply again and redraws the parts of it that have not changed on every update, cutting the data sent and the server work spe",
              "assets": 0
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/open-webui/open-webui/releases/tag/v0.11.1"
    },
    {
      "id": "10b8143111914c6b492f",
      "title": "huggingface/transformers Patch release: v5.15.1",
      "content_text": "Patch release v5.15.1\nThis patch most notably solves a few issues with DFlash and MTP candidate generators, as well as an issue where images could sometimes not be processed on accelerator if using Lanczos filter.\nIt contains the following commits:\nFix DFlash candidate token device mismatch with devicemap=\"auto\" (#47877) by @sywangyi and @Cyrilvallez\nAlign logit distributions for CandidateGenerators using sampling (#48007) by @Cyrilvallez\nFix MTP config when mlplayertypes is absent (#48015) by…",
      "date_published": "2026-08-19T10:50:47Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "10b8143111914c6b492f",
          "source": "github",
          "type": "change",
          "key": "https://github.com/huggingface/transformers/releases/tag/v5.15.1",
          "source_url": "https://api.github.com/repos/huggingface/transformers/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "huggingface/transformers Patch release: v5.15.1",
            "item_type": "release",
            "summary": "added: huggingface/transformers Patch release: v5.15.1",
            "after": {
              "repo": "huggingface/transformers",
              "title": "huggingface/transformers Patch release: v5.15.1",
              "tag": "v5.15.1",
              "url": "https://github.com/huggingface/transformers/releases/tag/v5.15.1",
              "published_at": "2026-08-19T10:50:47Z",
              "prerelease": false,
              "author": "Cyrilvallez",
              "summary": "Patch release v5.15.1\nThis patch most notably solves a few issues with DFlash and MTP candidate generators, as well as an issue where images could sometimes not be processed on accelerator if using Lanczos filter.\nIt contains the following commits:\nFix DFlash candidate token device mismatch with devicemap=\"auto\" (#47877) by @sywangyi and @Cyrilvallez\nAlign logit distributions for CandidateGenerators using sampling (#48007) by @Cyrilvallez\nFix MTP config when mlplayertypes is absent (#48015) by @Cyrilvallez\nFallback from 'lanczos' to 'bicubic' when on cuda (#48026) by @zucchini-nlp\nFix gemma4 video to device (#47896) by @guarin",
              "assets": 0
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/huggingface/transformers/releases/tag/v5.15.1"
    },
    {
      "id": "b65e8130fc3278de11bc",
      "title": "comfyanonymous/ComfyUI v0.33.1",
      "content_text": "What's Changed\nFix KSamplerAdvanced with addnoise disabled on nested latents by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/15447\nUpdate workflow templates to v0.11.40 by @comfyui-wiki in https://github.com/Comfy-Org/ComfyUI/pull/15522\nchore: replace api nodes -> partner nodes in README by @robinjhuang in https://github.com/Comfy-Org/ComfyUI/pull/15519\nFix PreviewAny escaping non-ASCII text in dict and list previews by @christian-byrne in…",
      "date_published": "2026-08-13T22:18:34Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "b65e8130fc3278de11bc",
          "source": "github",
          "type": "change",
          "key": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.33.1",
          "source_url": "https://api.github.com/repos/comfyanonymous/ComfyUI/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "comfyanonymous/ComfyUI v0.33.1",
            "item_type": "release",
            "summary": "added: comfyanonymous/ComfyUI v0.33.1",
            "after": {
              "repo": "comfyanonymous/ComfyUI",
              "title": "comfyanonymous/ComfyUI v0.33.1",
              "tag": "v0.33.1",
              "url": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.33.1",
              "published_at": "2026-08-13T22:18:34Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "What's Changed\nFix KSamplerAdvanced with addnoise disabled on nested latents by @kijai in https://github.com/Comfy-Org/ComfyUI/pull/15447\nUpdate workflow templates to v0.11.40 by @comfyui-wiki in https://github.com/Comfy-Org/ComfyUI/pull/15522\nchore: replace api nodes -> partner nodes in README by @robinjhuang in https://github.com/Comfy-Org/ComfyUI/pull/15519\nFix PreviewAny escaping non-ASCII text in dict and list previews by @christian-byrne in https://github.com/Comfy-Org/ComfyUI/pull/15513\nFix float64 device in ltx diffusion decoder. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15516\nSupport anima tunes with extra blocks. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15555\nDon't disable dynamic vram on WSL. by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15562\nQuery pytorch for aotriton support instead of listing its lib directory by @glop102 in https://github.com/Comfy-Org/ComfyUI/pull/15412\nFix Generate Text ignoring thinking=false on Gemma4 E2B/E4B by @DrJKL in https://github.com/Comfy-Org/ComfyUI/pull/15278\nUpdate comfy-kitchen package version to 0.2.31 by @comfyanonymous in https://github.com/Comfy-Org/ComfyUI/pull/15564\n[Partner Nodes] feat(MiniMax): add ContextIR and Regenerate To 2K nodes by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15471\nImplement Minimax Music 3 + Core Support for Cuda Graphs by @rattus128 in https://github.com/Comfy-Org/ComfyUI/pull/15570\n[Partner Nodes] feat(Bria): add GenFill, Eraser, Expand and Increase Resolution nodes by @bigcat88 in https://github.com/Comfy-Org/ComfyUI/pull/15572\nllama: fix non-local x path by @rattus128 in https://github.com/Comfy-Org/ComfyUI/pull/15580\nUpdate workflow templates to v0.11.41 by @comfyui-wiki in https://github.com/Comfy-Org/ComfyUI/pull/15578\nminimax: early detect qkv vs q,k,v by @rattus128 in https://github.com/Comfy-Org/ComfyUI/pull/15581\nFix minimax music not working on non dynamic vram. https://github.com/Comfy-Org/ComfyUI/pull",
              "assets": 4
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/Comfy-Org/ComfyUI/releases/tag/v0.33.1"
    },
    {
      "id": "c8369cbfd12301518036",
      "title": "vllm-project/vllm v0.27.1",
      "content_text": "This is a patch release on top of v0.27.0.\nSupport quantized DSpark Markov heads (#50424)",
      "date_published": "2026-08-11T10:47:49Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "c8369cbfd12301518036",
          "source": "github",
          "type": "change",
          "key": "https://github.com/vllm-project/vllm/releases/tag/v0.27.1",
          "source_url": "https://api.github.com/repos/vllm-project/vllm/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "vllm-project/vllm v0.27.1",
            "item_type": "release",
            "summary": "added: vllm-project/vllm v0.27.1",
            "after": {
              "repo": "vllm-project/vllm",
              "title": "vllm-project/vllm v0.27.1",
              "tag": "v0.27.1",
              "url": "https://github.com/vllm-project/vllm/releases/tag/v0.27.1",
              "published_at": "2026-08-11T10:47:49Z",
              "prerelease": false,
              "author": "khluu",
              "summary": "This is a patch release on top of v0.27.0.\nSupport quantized DSpark Markov heads (#50424)",
              "assets": 8
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/vllm-project/vllm/releases/tag/v0.27.1"
    },
    {
      "id": "1050f2ae8765a9b8a525",
      "title": "vllm-project/vllm v0.27.0",
      "content_text": "vLLM v0.27.0 Release Notes\nHighlights\nThis release features 561 commits from 242 contributors (64 new)!\nKimi K3 support with a full stack landing in one release: core model files and kernels (#50089, #50000), Python (#50093) and Rust (#50104) frontends, AttnRes kernels (#50090), DeepGEMM support (#50458), compressed-tensors quantized checkpoints (#50500), DSpark AR fusion (#50242), and an option to shard the shared expert instead of replicating it (#50656).\nMore new models: Qwen3.5 text-only…",
      "date_published": "2026-08-10T21:18:11Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "1050f2ae8765a9b8a525",
          "source": "github",
          "type": "change",
          "key": "https://github.com/vllm-project/vllm/releases/tag/v0.27.0",
          "source_url": "https://api.github.com/repos/vllm-project/vllm/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "vllm-project/vllm v0.27.0",
            "item_type": "release",
            "summary": "added: vllm-project/vllm v0.27.0",
            "after": {
              "repo": "vllm-project/vllm",
              "title": "vllm-project/vllm v0.27.0",
              "tag": "v0.27.0",
              "url": "https://github.com/vllm-project/vllm/releases/tag/v0.27.0",
              "published_at": "2026-08-10T21:18:11Z",
              "prerelease": false,
              "author": "khluu",
              "summary": "vLLM v0.27.0 Release Notes\nHighlights\nThis release features 561 commits from 242 contributors (64 new)!\nKimi K3 support with a full stack landing in one release: core model files and kernels (#50089, #50000), Python (#50093) and Rust (#50104) frontends, AttnRes kernels (#50090), DeepGEMM support (#50458), compressed-tensors quantized checkpoints (#50500), DSpark AR fusion (#50242), and an option to shard the shared expert instead of replicating it (#50656).\nMore new models: Qwen3.5 text-only dense and MoE models (#50210) with EVS video token pruning (#48912), K-EXAONE-2.0-750B-A37B (#50524), VaultGemma via the Transformers modeling backend (#49803), and jina-embeddings-v5-text-nano (#50688).\nPyTorch 2.13.0 upgrade along with torchvision 0.28.0 and Triton 3.7.1 (#48155) — this is a breaking environment change; XPU (#48677) and CPU (#50412) followed to torch 2.13 as well.\nFlashAttention 4 integration deepens on SM100: FP8 KV cache support (#42569) and headdim-256 support (#42669), backed by a new JIT warmup infrastructure (#47451) and runner-owned Triton kernel warmup (#49903) that remove first-request compilation stalls.\nDeepSeek-V4 performance push: sequence parallelism (#46789), 2x kernel improvement by skipping empty c128 launches (#48957), 3.4% E2E TTFT from skipping unneeded topk/router (#49486), 3.9% E2E TTFT from workspace reuse (#49236), 1.88x kernel from removing a redundant full kernel (#50298), adaptive topk width (1.0% E2E, #50004), 448 MiB GPU memory saved in the PP buffer (#50312), a compact MXFP4 indexer KV cache (#48993), and removal of sparse-MLA q-head padding on FlashInfer >= 0.6.14 (#48047).\nModel Runner V2 expands to non-generative workloads: encoder-only attention (#49331), sequence pooling for embedding/classification (#48791), encoder token classification (#50293) and token embedding (#50574), BGE-M3 pooling (#50661), multimodal on CPU (#50073), a multi-layer MTP speculator (#48892), and PCP now selects MRV2 (#50034).\nResilient large-scale ser",
              "assets": 8
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/vllm-project/vllm/releases/tag/v0.27.0"
    },
    {
      "id": "5bc225903c9d4dbe9d7c",
      "title": "huggingface/transformers Release: v5.15.0",
      "content_text": "Release v5.15.0\nNew Model additions\nMeta Muse Glimmer\nMuse Glimmer, released today, is Meta’s new multimodal model, especially designed for agentic use cases. Distilled from Muse to 30B parameters, and released under the Apache 2.0 license, it can be deployed to local setups for privacy-aware applications such as coding, document analysis, personal assistants, Claw- or Hermes-like setups.\nMuse Glimmer is a dense 30B parameter model consisting of:\n2B ViT-style encoder for vision (Perception…",
      "date_published": "2026-08-10T10:28:13Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "5bc225903c9d4dbe9d7c",
          "source": "github",
          "type": "change",
          "key": "https://github.com/huggingface/transformers/releases/tag/v5.15.0",
          "source_url": "https://api.github.com/repos/huggingface/transformers/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "huggingface/transformers Release: v5.15.0",
            "item_type": "release",
            "summary": "added: huggingface/transformers Release: v5.15.0",
            "after": {
              "repo": "huggingface/transformers",
              "title": "huggingface/transformers Release: v5.15.0",
              "tag": "v5.15.0",
              "url": "https://github.com/huggingface/transformers/releases/tag/v5.15.0",
              "published_at": "2026-08-10T10:28:13Z",
              "prerelease": false,
              "author": "LysandreJik",
              "summary": "Release v5.15.0\nNew Model additions\nMeta Muse Glimmer\nMuse Glimmer, released today, is Meta’s new multimodal model, especially designed for agentic use cases. Distilled from Muse to 30B parameters, and released under the Apache 2.0 license, it can be deployed to local setups for privacy-aware applications such as coding, document analysis, personal assistants, Claw- or Hermes-like setups.\nMuse Glimmer is a dense 30B parameter model consisting of:\n2B ViT-style encoder for vision (Perception Encoder)\n28B parameter text decoder\nWe're covering it in the following blogpost: http://hf.co/blog/muse-glimmer\n---\nGraniteMoeSWA & GraniteSWA\nLinks: Documentation\nAdd Granite-swa and Granitemoe-swa model support (#47179) by @daviswer in #47179\nLinks: Documentation\nAdd Granite-swa and Granitemoe-swa model support (#47179) by @daviswer in #47179\n---\nA.X-K1 & A.X-K2\nLinks: Documentation\nAdd AXK2 from SKT (#47528) by @vasqu in #47528\nLinks: Documentation\naddaxk1 (#46867) by @kmswin1 in #46867\n---\nCosmos3 Edge\nLinks: Documentation\nAdd Cosmos3 Edge model support (#47181) by @atharvajoshi10 in #47181\nBreaking changes\nKernels are now opt-in rather than mandatory for linear attention models (Mamba, GDN, Conv-only, etc.), so users who relied on automatic kernel selection must explicitly enable kernels to maintain previous behavior.\n🚨 [Kernels] Refactor all linear attn models & native kernels fallback (#47630) by @vasqu\nThe cache cropping API now only accepts negative values (relative offsets) instead of absolute sizes, so users calling crop methods directly must update their code to pass negative values accordingly.\n🚨 [cache] Cropping can only be done with negative values (#47720) by @Cyrilvallez\nT5 and its model family (MT5, LongT5, etc.) now support SDPA and other attention backends via ALLATTENTIONFUNCTIONS, meaning the default attention implementation may change and users relying on the previous eager-only path should explicitly set attnimplementation=\"eager\" if needed.\n🚨 Enable SDPA (",
              "assets": 0
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/huggingface/transformers/releases/tag/v5.15.0"
    },
    {
      "id": "e1ba6959e9cc92912402",
      "title": "open-webui/open-webui v0.11.0",
      "content_text": "Added\n🎨 Redesigned interface. Open WebUI has been visually rebuilt from the ground up. All aspects of the User Interface, from the chat view to the admin panel. Now with a narrower conversation column, lighter typography, tidier spacing, consistent menus and dropdowns, clearly outlined text boxes, and settings rearranged. Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit…",
      "date_published": "2026-07-27T09:30:15Z",
      "_unlimitedpipe": {
        "event": {
          "schema": "unlimitedpipe.event/1",
          "id": "e1ba6959e9cc92912402",
          "source": "github",
          "type": "change",
          "key": "https://github.com/open-webui/open-webui/releases/tag/v0.11.0",
          "source_url": "https://api.github.com/repos/open-webui/open-webui/releases",
          "timestamp": null,
          "observed_at": "2026-09-24T10:12:41Z",
          "data": {
            "change": "added",
            "label": "open-webui/open-webui v0.11.0",
            "item_type": "release",
            "summary": "added: open-webui/open-webui v0.11.0",
            "after": {
              "repo": "open-webui/open-webui",
              "title": "open-webui/open-webui v0.11.0",
              "tag": "v0.11.0",
              "url": "https://github.com/open-webui/open-webui/releases/tag/v0.11.0",
              "published_at": "2026-07-27T09:30:15Z",
              "prerelease": false,
              "author": "github-actions[bot]",
              "summary": "Added\n🎨 Redesigned interface. Open WebUI has been visually rebuilt from the ground up. All aspects of the User Interface, from the chat view to the admin panel. Now with a narrower conversation column, lighter typography, tidier spacing, consistent menus and dropdowns, clearly outlined text boxes, and settings rearranged. Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, Commit, #27178, Commit, Commit\n🤖 Sub-agents. Administrators can now enable sub-agents, which let a model hand parts of a task to background helper agents that run their own tool-driven conversations and report results back into the chat, tuned through new \"ENABLESUBAGENTS\", concurrency, iteration, and system-prompt settings. Commit, Commit, Commit, Commit\n📂 Folder pages. Opening a folder now takes you to its own page, where its chats load a page at a time, can be sorted by title or last updated, and you can start a new chat straight from the folder. Commit\n⏲️ Chat timers. The assistant can now set a timer that brings a prompt back into the conversation later, after a delay or at a set time, and can drop it automatically if you read the chat or reply before it fires. Commit\n🔔 Notification targets. Notifications now have their own settings tab where you can send them to several webhook destinations, each picking which events it wants, from chats finishing or failing to channel messages and calendar alerts, with a test button and a choice between always notifying or only when you are away, and any webhook you already had is carried over for you. Commit, Commit, Commit, #24750\n🗯️ Full replies in channels. A reply from the assistant in a channel is now saved and shown in full, with its reasoning, tool calls and other structured parts, where it previously came through blank. Commit, #267",
              "assets": 0
            }
          },
          "metadata": {
            "method": "github-api"
          },
          "provenance": [
            {
              "step": "github",
              "version": "0.3.1"
            },
            {
              "step": "filter",
              "version": "0.3.1",
              "args": {
                "expr": "not (tag matches \"^b[0-9]+$\")"
              }
            },
            {
              "step": "map",
              "version": "0.3.1",
              "args": {
                "assign": [
                  "title=repo + \" \" + title"
                ]
              }
            },
            {
              "step": "sort",
              "version": "0.3.1",
              "args": {
                "by": [
                  "published_at"
                ],
                "reverse": true
              }
            },
            {
              "step": "diff",
              "version": "0.3.1",
              "args": {
                "namespace": "ai-releases",
                "only": [
                  "added"
                ],
                "emit_initial": true
              }
            }
          ]
        }
      },
      "url": "https://github.com/open-webui/open-webui/releases/tag/v0.11.0"
    }
  ]
}