{"apiVersion":"v1","taxonomyVersion":1,"generatedAt":"2026-10-29T20:43:52.279Z","data":{"recordId":"wsa_jhgxqj8inf67","score":0.981,"confidence":"high","capped":false,"capReason":null,"sources":[{"n":1,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/v0.5.0","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-28T00:15:38.667Z","detectedPublishedAt":"2026-09-23T20:50:44.000Z","rights":"original"},{"n":2,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11163","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-09-24T19:44:21.674Z","detectedPublishedAt":"2026-09-24T19:30:47.000Z","rights":"original"},{"n":3,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11179","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-09-25T13:50:14.796Z","detectedPublishedAt":"2026-09-25T13:44:54.000Z","rights":"original"},{"n":4,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11223","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-09-28T02:07:13.899Z","detectedPublishedAt":"2026-09-27T22:06:22.000Z","rights":"original"},{"n":5,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11240","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-09-29T02:14:19.361Z","detectedPublishedAt":"2026-09-28T22:30:05.000Z","rights":"original"},{"n":6,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11298","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-09-30T20:28:24.949Z","detectedPublishedAt":"2026-09-30T20:22:47.000Z","rights":"original"},{"n":7,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11345","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-02T14:42:09.160Z","detectedPublishedAt":"2026-10-02T13:11:07.000Z","rights":"original"},{"n":8,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11368","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-03T02:46:06.465Z","detectedPublishedAt":"2026-10-03T02:38:39.000Z","rights":"original"},{"n":9,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11371","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-03T08:48:07.356Z","detectedPublishedAt":"2026-10-03T08:36:57.000Z","rights":"original"},{"n":10,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11400","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-05T03:02:11.446Z","detectedPublishedAt":"2026-10-05T00:06:48.000Z","rights":"original"},{"n":11,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11408","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-05T09:04:12.308Z","detectedPublishedAt":"2026-10-05T08:37:59.000Z","rights":"original"},{"n":12,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11507","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-08T15:33:15.947Z","detectedPublishedAt":"2026-10-08T15:31:21.000Z","rights":"original"},{"n":13,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11559","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-11T22:00:32.383Z","detectedPublishedAt":"2026-10-11T19:17:37.000Z","rights":"original"},{"n":14,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11695","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-16T16:41:37.619Z","detectedPublishedAt":"2026-10-16T11:16:03.000Z","rights":"original"},{"n":15,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11714","publisher":"GitHub watchlist","authority":"owner-controlled","verification":"verified","lastRetrievedAt":"2026-10-17T16:50:37.065Z","detectedPublishedAt":"2026-10-17T16:20:21.000Z","rights":"original"},{"n":16,"url":"https://github.com/ggml-org/llama.cpp/releases/tag/b11146","publisher":"GitHub watchlist","authority":"other","verification":"verified","lastRetrievedAt":"2026-10-28T00:15:38.667Z","detectedPublishedAt":"2026-09-23T18:40:01.000Z","rights":"original"}],"claims":[{"id":"4aeae410-f72d-49ae-84d8-5a4e3c3d8d4b","kind":"availability","statement":"The release is marked as a pre-release.","citations":[{"source":2,"supportProbability":0.95,"supports":true,"quote":null},{"source":16,"supportProbability":0.95,"supports":true,"quote":null}]},{"id":"a45ae7c7-3add-47ee-908d-522521f5ca30","kind":"availability","statement":"Builds are listed for macOS, iOS, Linux, Android, Windows, and openEuler.","citations":[{"source":2,"supportProbability":0.81,"supports":true,"quote":null},{"source":16,"supportProbability":0.76,"supports":true,"quote":null}]},{"id":"31d109bd-521c-4d11-95b7-cb3996e131be","kind":"capability","statement":"The batch API supports token, embedding, and state inputs.","citations":[{"source":2,"supportProbability":0.87,"supports":true,"quote":null}]},{"id":"cf7be060-9812-449d-87b7-8e87e97261fb","kind":"capability","statement":"The release adds the llama_batch_ext API.","citations":[{"source":2,"supportProbability":0.92,"supports":true,"quote":{"text":"llama: add llama_batch_ext ( #24669 )","rights":"original"}}]},{"id":"924a95a6-9470-4805-87c0-5b68d854da06","kind":"availability","statement":"b11179 is marked as a pre-release.","citations":[{"source":3,"supportProbability":0.96,"supports":true,"quote":null}]},{"id":"36606186-8c0c-41bb-9851-8c952f4843fe","kind":"deployment","statement":"The listed builds include CPU and accelerator-specific variants.","citations":[{"source":3,"supportProbability":0.25,"supports":false,"quote":null}]},{"id":"fbfc20a7-917f-4733-b339-d0c89a19f32a","kind":"availability","statement":"The release lists builds for macOS, iOS, Linux, Android, and Windows.","citations":[{"source":3,"supportProbability":0.74,"supports":true,"quote":null},{"source":16,"supportProbability":0.77,"supports":true,"quote":null}]},{"id":"7af83320-2b17-4155-abbf-c100cf7b6fde","kind":"capability","statement":"The release adds support for sparse FA on SYCL.","citations":[{"source":3,"supportProbability":0.91,"supports":true,"quote":null}]},{"id":"485e71d3-ed34-4ff6-8d75-650b6cdd5251","kind":"deployment","statement":"The release lists builds for macOS, iOS, Linux, Android, Windows, and openEuler.","citations":[{"source":4,"supportProbability":0.91,"supports":true,"quote":null},{"source":16,"supportProbability":0.77,"supports":true,"quote":null}]},{"id":"c6dda7d3-117b-4e00-9b17-d2c210d8e1ea","kind":"capability","statement":"The server now permits chunked prefill for RANK pooling when the context is causal.","citations":[{"source":4,"supportProbability":0.96,"supports":true,"quote":{"text":"can_split() now permits chunked prefill for RANK pooling when the","rights":"original"}}]},{"id":"d450840c-8247-4038-a7bb-b490f5bec05f","kind":"capability","statement":"The change addresses long-document and multimodal reranking for causal models including Qwen3 and Qwen3-VL.","citations":[{"source":4,"supportProbability":0.92,"supports":true,"quote":null}]},{"id":"064575c3-863f-4794-916d-9d038a320b3c","kind":"availability","statement":"Release b11223 is marked as a pre-release.","citations":[{"source":4,"supportProbability":0.97,"supports":true,"quote":null}]},{"id":"856a79aa-e302-4cfc-aba1-c0b26a3df63f","kind":"capability","statement":"Wrapped content arrays are accepted for multimodal embedding requests.","citations":[{"source":5,"supportProbability":0.96,"supports":true,"quote":{"text":"Accept the OpenAI-style wrapped content array format for multimodal","rights":"original"}}]},{"id":"16208e4d-c025-4a6a-b15f-6036b500568a","kind":"capability","statement":"The server supports multimodal input for Qwen3-VL-Embedding.","citations":[{"source":5,"supportProbability":0.96,"supports":true,"quote":{"text":"server : support multimodal input for /v1/embeddings (Qwen3-VL-Embedding)","rights":"original"}}]},{"id":"545c9a19-a65a-4a7f-8948-9df038cc30f4","kind":"capability","statement":"The server supports typed vision, audio, and video content input for the /v1/embeddings endpoint.","citations":[{"source":5,"supportProbability":0.96,"supports":true,"quote":{"text":"server : support typed content (vision/audio/video) input for /v1/embeddings endpoint ( #29556 )","rights":"original"}}]},{"id":"7ccbce6c-bd1e-4da3-8080-e176b4d93213","kind":"availability","statement":"The GitHub release is marked Pre-release.","citations":[{"source":5,"supportProbability":0.96,"supports":true,"quote":null},{"source":16,"supportProbability":0.97,"supports":true,"quote":null}]},{"id":"dd4a0c46-ace6-4b27-8c62-e1879bde1cdc","kind":"capability","statement":"Bare content arrays are rejected with a migration message.","citations":[{"source":5,"supportProbability":0.99,"supports":true,"quote":{"text":"Bare content arrays (the unwrapped shape) are","rights":"original"}}]},{"id":"86505b2e-59dd-4e2e-b0dd-ae6555874a1a","kind":"capability","statement":"Legacy formats continue to work through tokenize_input_prompts.","citations":[{"source":5,"supportProbability":0.97,"supports":true,"quote":null}]},{"id":"c882cbb4-5e4a-4768-89bb-61395fe15698","kind":"capability","statement":"The release adds MiMo support for dflash conversion and feature extraction.","citations":[{"source":6,"supportProbability":0.9,"supports":true,"quote":null}]},{"id":"1c5fb59d-3d34-40e5-a171-9f003e49bd19","kind":"availability","statement":"The GitHub release is marked Pre-release.","citations":[{"source":6,"supportProbability":0.96,"supports":true,"quote":null},{"source":16,"supportProbability":0.97,"supports":true,"quote":null}]},{"id":"f2166412-573b-4deb-9d60-bc2dce7c9ed2","kind":"availability","statement":"The release lists builds for macOS, Linux, Android, and Windows.","citations":[{"source":6,"supportProbability":0.74,"supports":true,"quote":null},{"source":16,"supportProbability":0.77,"supports":true,"quote":null}]},{"id":"c1f8d8b2-c68a-4631-81f0-e7598d8f4176","kind":"deployment","statement":"The release lists Linux arm64 builds for Snapdragon CPU, Adreno GPU and Hexagon NPU.","citations":[{"source":7,"supportProbability":0.96,"supports":true,"quote":{"text":"Linux arm64 (Snapdragon: CPU, Adreno GPU, Hexagon NPU) - setup guide","rights":"original"}},{"source":16,"supportProbability":0.94,"supports":true,"quote":null}]},{"id":"2514ad0a-4088-47d0-bb1f-a35040f246c5","kind":"capability","statement":"The release adds q2_k and q3_k quant type support for Hexagon.","citations":[{"source":7,"supportProbability":0.96,"supports":true,"quote":{"text":"hexagon: add q2_k and q3_k quant type support ( #29717 )","rights":"original"}}]},{"id":"477045de-5edd-41da-9f93-5ffdbee76a3d","kind":"availability","statement":"Release b11345 is marked as a pre-release.","citations":[{"source":7,"supportProbability":0.97,"supports":true,"quote":null}]},{"id":"2ad71284-ff3d-4a3f-b9d3-68275cbf571a","kind":"availability","statement":"The release lists builds for macOS, iOS, Linux, Android, Windows and openEuler.","citations":[{"source":7,"supportProbability":0.86,"supports":true,"quote":null},{"source":16,"supportProbability":0.78,"supports":true,"quote":null}]},{"id":"421c1a1f-2e80-4740-81cb-7ec1099fa2a4","kind":"deployment","statement":"The release lists Android arm64 builds for Snapdragon CPU, Adreno GPU and Hexagon NPU.","citations":[{"source":7,"supportProbability":0.95,"supports":true,"quote":{"text":"Android arm64 (Snapdragon: CPU, Adreno GPU, Hexagon NPU) - setup guide","rights":"original"}},{"source":16,"supportProbability":0.96,"supports":true,"quote":null}]},{"id":"e50902d3-cb4e-47de-a21a-9bad9b95e1cf","kind":"capability","statement":"The target verifies drafts by rejection sampling.","citations":[{"source":8,"supportProbability":0.96,"supports":true,"quote":{"text":"Make the drafter probabilistic and the target verify by rejection sampling","rights":"original"}}]},{"id":"c515a7f1-3ead-4813-8275-2c6185454704","kind":"capability","statement":"The release adds probabilistic sampling for simple draft and MTP.","citations":[{"source":8,"supportProbability":0.96,"supports":true,"quote":{"text":"spec : add probabilistic sampling for simple draft and MTP ( #27694 )","rights":"original"}}]},{"id":"b11a2fe1-94ac-4813-a03e-a284359d7c47","kind":"capability","statement":"Probabilistic draft sampling is controlled by a flag whose default value is greedy.","citations":[{"source":8,"supportProbability":0.97,"supports":true,"quote":{"text":"Fallback to argmax sampling for grammar-constrained requests and adding flag for enabling probabilistic draft sampling.","rights":"original"}}]},{"id":"fef0dd49-660a-41d4-b4fc-f67a5d07e86b","kind":"availability","statement":"The release is marked pre-release and was released on 03 Oct.","citations":[{"source":8,"supportProbability":0.96,"supports":true,"quote":null}]},{"id":"eb227ef4-d564-4549-b32a-78b67d94d13f","kind":"capability","statement":"Rejection sampling supports grammar-constrained requests.","citations":[{"source":8,"supportProbability":0.95,"supports":true,"quote":{"text":"Support grammar-constrained requests in rejection sampling","rights":"original"}}]},{"id":"cf794979-9675-4f66-ac63-a1bb5775df66","kind":"availability","statement":"The GitHub release is marked Pre-release.","citations":[{"source":9,"supportProbability":0.95,"supports":true,"quote":null},{"source":16,"supportProbability":0.96,"supports":true,"quote":null}]},{"id":"68e1dce3-0f8f-4044-b204-c41f242368f5","kind":"availability","statement":"The release lists platform builds for macOS, iOS, Linux, Android, and Windows.","citations":[{"source":9,"supportProbability":0.76,"supports":true,"quote":null},{"source":16,"supportProbability":0.73,"supports":true,"quote":null}]},{"id":"eb62c906-4deb-47b2-b407-03c9baa5c9d8","kind":"capability","statement":"The release adds initial support for the Clef decision model.","citations":[{"source":9,"supportProbability":0.86,"supports":true,"quote":{"text":"model: add support for clef decision model (text-only) ( #29831 )","rights":"original"}}]},{"id":"a2672860-e67a-4a4e-8f97-06de4c48bf6f","kind":"capability","statement":"The Clef support is text-only.","citations":[{"source":9,"supportProbability":0.94,"supports":true,"quote":{"text":"model: add support for clef decision model (text-only) ( #29831 )","rights":"original"}}]},{"id":"5e7edfd2-7c9e-46a3-af3f-151eb3082adf","kind":"availability","statement":"The release lists builds for macOS Apple Silicon and Intel, and an iOS XCFramework.","citations":[{"source":10,"supportProbability":0.96,"supports":true,"quote":null},{"source":16,"supportProbability":0.96,"supports":true,"quote":null}]},{"id":"40b0e883-8824-4b07-8a12-79edf760caf9","kind":"availability","statement":"b11400 is marked as a pre-release.","citations":[{"source":10,"supportProbability":0.97,"supports":true,"quote":null}]},{"id":"d80cf4d8-71cf-462d-9c34-9f9218d3dfb2","kind":"capability","statement":"The release adds support for both embeddings and raw tokens in a batch.","citations":[{"source":10,"supportProbability":0.96,"supports":true,"quote":null}]},{"id":"05b6a98a-2955-4c03-a0b3-1a471c76e141","kind":"availability","statement":"The release lists Linux builds for CPU, Vulkan, CUDA, ROCm, OpenVINO and SYCL configurations.","citations":[{"source":10,"supportProbability":0.86,"supports":true,"quote":null},{"source":16,"supportProbability":0.85,"supports":true,"quote":null}]},{"id":"547414f1-b26b-4d94-ae86-ee5beeb57099","kind":"availability","statement":"The release lists Android and Windows builds.","citations":[{"source":10,"supportProbability":0.82,"supports":true,"quote":null},{"source":16,"supportProbability":0.82,"supports":true,"quote":null}]},{"id":"24545a98-7011-4a4a-a91f-534af86f1544","kind":"recommendation-relevant","statement":"On that test, pp128 throughput increased from 10.70 to 93.87 tokens per second.","citations":[{"source":11,"supportProbability":0.96,"supports":true,"quote":null}]},{"id":"ad59f295-a4ca-4b46-a092-27fdc6354a79","kind":"recommendation-relevant","statement":"The release reports Qwen2.5-0.5B-Instruct Q8_0 tested on a Milk-V Jupiter.","citations":[{"source":11,"supportProbability":0.97,"supports":true,"quote":null}]},{"id":"fc591d52-5beb-43e3-b862-45c72793b487","kind":"availability","statement":"b11408 is marked as a pre-release on GitHub.","citations":[{"source":11,"supportProbability":0.96,"supports":true,"quote":null}]},{"id":"ffa64fa1-3b55-4549-a029-63839a9b48fc","kind":"capability","statement":"The release marks Q8_0 as supported on the X60.","citations":[{"source":11,"supportProbability":0.87,"supports":true,"quote":null}]},{"id":"b0c6c265-7581-454b-adc8-6d25c010b74f","kind":"capability","statement":"b11408 adds a Q8_0 IME1 matrix kernel for the SpacemiT X60.","citations":[{"source":11,"supportProbability":0.98,"supports":true,"quote":{"text":"ggml-cpu : add Q8_0 IME1 matrix kernel for SpacemiT X60 ( #28479 )","rights":"original"}}]},{"id":"43116094-30fa-4481-b9d8-2e592530aa90","kind":"availability","statement":"b11507 is marked as a pre-release.","citations":[{"source":12,"supportProbability":0.95,"supports":true,"quote":null}]},{"id":"ec2ce369-03b0-4cfc-80b9-6b3c30dbe9ab","kind":"availability","statement":"The release lists macOS, iOS, Linux, Android, and Windows builds.","citations":[{"source":12,"supportProbability":0.77,"supports":true,"quote":null},{"source":16,"supportProbability":0.8,"supports":true,"quote":null}]},{"id":"d765dd5c-fa0a-4dae-973c-a385749b2fdb","kind":"recommendation-relevant","statement":"Linux and Windows builds are listed for CUDA 12 and CUDA 13.","citations":[{"source":12,"supportProbability":0.63,"supports":false,"quote":null},{"source":16,"supportProbability":0.72,"supports":true,"quote":null}]},{"id":"aa2776d9-14d2-4832-8f65-d714f2ba2ba4","kind":"capability","statement":"b11507 adds support for MoE cache over multiple GPUs.","citations":[{"source":12,"supportProbability":0.94,"supports":true,"quote":{"text":"llama: support MoE cache over multiple GPUs ( #30112 )","rights":"original"}}]},{"id":"ee6e87ad-224f-4990-a3b8-cac0cbefdc91","kind":"capability","statement":"The release improves OpenCL support for FA and DK512 for Gemma-4.","citations":[{"source":13,"supportProbability":0.93,"supports":true,"quote":null}]},{"id":"97661d80-2e25-4204-9ed6-a3cdc560c774","kind":"deployment","statement":"The release lists builds for macOS, iOS, Linux, Android, Windows, and openEuler.","citations":[{"source":13,"supportProbability":0.8,"supports":true,"quote":null},{"source":16,"supportProbability":0.78,"supports":true,"quote":null}]},{"id":"0ddb98a1-ec78-42c1-a3b2-8427d33387ce","kind":"availability","statement":"The release is marked Pre-release.","citations":[{"source":13,"supportProbability":0.95,"supports":true,"quote":null},{"source":16,"supportProbability":0.95,"supports":true,"quote":null}]},{"id":"3eae7eac-9053-4b78-9663-6be2d459b0e9","kind":"capability","statement":"The release enables Gemma-4 E4B GPU decode through OpenCL.","citations":[{"source":13,"supportProbability":0.93,"supports":true,"quote":{"text":"opencl: enable Gemma-4 E4B GPU decode","rights":"original"}}]},{"id":"236288b0-2b65-4f76-a7ea-a00929f8d3d7","kind":"capability","statement":"The release extends Gemma-4 GPU decode through OpenCL to E2B.","citations":[{"source":13,"supportProbability":0.93,"supports":true,"quote":{"text":"opencl: enable Gemma-4 E4B GPU decode","rights":"original"}}]},{"id":"a58c2c23-47d5-46e3-b248-a32b054a5fc8","kind":"capability","statement":"The release adds MTP support for Mellum.","citations":[{"source":14,"supportProbability":0.92,"supports":true,"quote":{"text":"model : MTP support for Mellum2.1 ( #30612 )","rights":"original"}}]},{"id":"802ec06e-2590-4ac3-a682-bd8fc1cc2727","kind":"capability","statement":"The release adds MTP support for Mellum2.1.","citations":[{"source":14,"supportProbability":0.93,"supports":true,"quote":{"text":"model : MTP support for Mellum2.1 ( #30612 )","rights":"original"}}]},{"id":"0ef9fa3b-0cc1-44e3-b482-43436b6d6663","kind":"capability","statement":"The release adds inline RoPE construction in Mellum.","citations":[{"source":14,"supportProbability":0.92,"supports":true,"quote":{"text":"model : inline rope construction in Mellum","rights":"original"}}]},{"id":"4ea2bea0-086d-4476-860a-cb04c312745e","kind":"availability","statement":"Builds are listed for macOS/iOS, Linux, Android, and Windows.","citations":[{"source":14,"supportProbability":0.74,"supports":true,"quote":null},{"source":16,"supportProbability":0.73,"supports":true,"quote":null}]},{"id":"4753c59c-93d7-4505-bfca-c78a4179832c","kind":"availability","statement":"b11695 is marked as a pre-release.","citations":[{"source":14,"supportProbability":0.95,"supports":true,"quote":null}]},{"id":"bb20f8de-96c1-4926-b6a1-27e419d0c1a9","kind":"availability","statement":"b11714 is marked as a pre-release.","citations":[{"source":15,"supportProbability":0.86,"supports":true,"quote":null}]},{"id":"5657efde-8280-43fd-8a6e-fca9e077dc89","kind":"capability","statement":"The Discover Models surface includes a curated catalog list with search and a details pane.","citations":[{"source":15,"supportProbability":0.84,"supports":true,"quote":{"text":"The Discover surface: the curated-catalog list with its search and","rights":"original"}}]},{"id":"f5cd7e12-130a-4b1d-94e7-04a665cbc2a1","kind":"capability","statement":"The Discover interface includes download options and a standalone download progress bar.","citations":[{"source":15,"supportProbability":0.78,"supports":true,"quote":{"text":"download options, and the standalone download progress bar that replaces","rights":"original"}}]},{"id":"a19c64e4-50cb-4747-ae36-78c634be1f53","kind":"capability","statement":"Paused downloads can be resumed, and incomplete downloads are remembered across restarts.","citations":[{"source":15,"supportProbability":0.75,"supports":true,"quote":null}]},{"id":"5100f26a-e5fb-4c2e-8029-6596b1264791","kind":"capability","statement":"The Discover setting gates browsing and downloading Hugging Face GGUF models.","citations":[{"source":15,"supportProbability":0.8,"supports":true,"quote":{"text":"The flag gates browsing and downloading HuggingFace GGUF models, so it is","rights":"original"}}]},{"id":"76e2c4c1-75d8-4de6-960d-bb69ad3046af","kind":"availability","statement":"The release is v0.5.0, published on 23 Sep.","citations":[{"source":1,"supportProbability":0.96,"supports":true,"quote":{"text":"The release expands hyper-connection, flash-attention, and fused MoE/SSM support across backends, with robustness, quantization, data-layout, and RPC/meta improvements.","rights":"original"}},{"source":16,"supportProbability":0.84,"supports":true,"quote":null}]},{"id":"0d5a8ffc-f82f-42af-957b-63cf711c6af2","kind":"capability","statement":"The release accelerates CUDA conv2d with implicit GEMM.","citations":[{"source":1,"supportProbability":0.98,"supports":true,"quote":{"text":"Accelerate CUDA conv2d with implicit GEMM ( #29135 )","rights":"original"}}]},{"id":"b78fc606-0fa3-4f6c-99c5-96e8d9283c46","kind":"capability","statement":"The release adds HRM-Text (DFM Mimir 1B) support.","citations":[{"source":1,"supportProbability":0.99,"supports":true,"quote":{"text":"It adds HRM-Text (DFM Mimir 1B) support, MiMo-V2.6 and HunyuanOCR conversion support, ggml 0.25.0 backend improvements, multi-address HTTP binding, image outputs from function calls, and several chat parser/UI fixes.","rights":"original"}}]},{"id":"4863785f-685d-49fd-9844-408d1f7bff96","kind":"capability","statement":"Server function-call outputs support input_image.","citations":[{"source":1,"supportProbability":0.97,"supports":true,"quote":{"text":"Add input_image support to server function-call outputs ( #22575 )","rights":"original"}}]},{"id":"bd413c53-cb7b-43d7-924d-f60eae850ded","kind":"capability","statement":"The server can bind to multiple comma-separated TCP addresses and UNIX sockets.","citations":[{"source":1,"supportProbability":0.95,"supports":true,"quote":{"text":"Allow the server to bind to multiple addresses ( #28690 )","rights":"original"}}]},{"id":"63dce2b5-b5d5-46d5-92a6-db5130253f56","kind":"capability","statement":"The release adds conversion support for MiMo-V2.6.","citations":[{"source":1,"supportProbability":0.98,"supports":true,"quote":{"text":"It adds HRM-Text (DFM Mimir 1B) support, MiMo-V2.6 and HunyuanOCR conversion support, ggml 0.25.0 backend improvements, multi-address HTTP binding, image outputs from function calls, and several chat parser/UI fixes.","rights":"original"}}]}]},"nextCursor":null,"license":{"catalog":{"name":"CC BY 4.0","spdx":"CC-BY-4.0","url":"https://creativecommons.org/licenses/by/4.0/","attribution":"wowsuchai","attributionUrl":"https://wowsuchai.withglint.app","covers":["catalog metadata","classifications","generated summaries","feed event metadata"]},"excluded":{"covers":["source quotations","linked source content"],"terms":"Not licensed by wowsuchai. Source quotations and linked source content keep their original rights."}}}