{"updatedAt":"2026-11-05 13:12","timezone":"Asia/Shanghai","tags":["全部","LLM","MoE","NVFP4","GGUF","多模态","边缘","ASR"],"items":[{"id":"hf-ripple-65b-preview","time":"11-05 12:50","slot":"digest","tag":"LLM","title":"ricursiveAI/Ripple-65B-Preview","why":"11-04 预览发布；可关注同系 NVFP4 / GGUF 部署旁路","params":"≈64.8B · ricursive · BF16 · ctx 2M · Apache-2.0 · pipeline 未标","source":"Hugging Face","href":"https://huggingface.co/ricursiveAI/Ripple-65B-Preview","accent":1,"pipeline_tag":null,"license":"apache-2.0","downloads":0,"architecture":"ricursive","dtype":"BF16","context_length":2097152,"createdAt":"2026-11-04T02:41:21.000Z","published":"上线 2026-11-04","releasedAt":"2026-11-04T02:42:18.000Z"},{"id":"hf-ista-deepseek-gsq-rco-gguf","time":"11-05 12:50","slot":"digest","tag":"GGUF","title":"ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-Coder-GGUF","why":"GSQ/RCO 代码向本地 MoE 量化；GGUF 体积约 103GB（totalFileSize）","params":"参数数未标 · GGUF · MIT · 下载 73K · image-text-to-text · MoE/code","source":"Hugging Face","href":"https://huggingface.co/ISTA-DASLab/DeepSeek-V4.1-Flash-1017-GSQ-RCO-Coder-GGUF","accent":0,"pipeline_tag":"image-text-to-text","license":"mit","downloads":73282,"dtype":"GGUF","createdAt":"2026-11-02T13:52:10.000Z","published":"上线 2026-11-02","releasedAt":"2026-11-02T13:52:10.000Z"},{"id":"hf-nota-mistral-nvfp4-pruned20","time":"11-05 12:50","slot":"digest","tag":"NVFP4","title":"nota-ai/Mistral-Large-4-1T-A52B-NVFP4-Global-Pruned-20","why":"NVFP4 + 20% expert pruning；gated；面向 vLLM 部署","params":"参数未公开 · NVFP4 · MoE pruning · license:other · gated · vLLM · 下载 1","source":"Hugging Face","href":"https://huggingface.co/nota-ai/Mistral-Large-4-1T-A52B-NVFP4-Global-Pruned-20","accent":2,"pipeline_tag":null,"license":"other","downloads":1,"dtype":"NVFP4","serving":["vLLM"],"gated":true,"createdAt":"2026-11-02T05:33:47.000Z","published":"上线 2026-11-02","releasedAt":"2026-11-02T05:33:47.000Z"},{"id":"hf-kimi-k31-pro","time":"11-05 12:50","slot":"digest","tag":"多模态","title":"moonshotai/Kimi-K3.1-Pro","why":"11-02 多模态上线，Together 可推理；Hub 参数统计含压缩张量","params":"Hub≈2.78T（含 U8 压缩）· kimi_k31 · 8-bit · ctx 1M · license:other · 下载 797 · Together","source":"Hugging Face","href":"https://huggingface.co/moonshotai/Kimi-K3.1-Pro","accent":2,"pipeline_tag":"image-text-to-text","license":"other","downloads":797,"architecture":"kimi_k31","dtype":"8-bit/compressed","context_length":1048576,"createdAt":"2026-11-02T04:13:51.000Z","published":"上线 2026-11-02","releasedAt":"2026-11-02T04:13:51.000Z"},{"id":"hf-nvidia-deepseek-v41-nvfp4","time":"11-05 12:50","slot":"digest","tag":"NVFP4","title":"nvidia/DeepSeek-V4.1-Flash-1017-NVFP4","why":"官方 NVFP4 量化部署变体，基于 DeepSeek-V4.1-Flash-1017","params":"Hub≈665B（量化张量）· deepseek_v41 · NVFP4/FP4 · MIT · 下载 3.6K · image-text-to-text","source":"Hugging Face","href":"https://huggingface.co/nvidia/DeepSeek-V4.1-Flash-1017-NVFP4","accent":1,"pipeline_tag":"image-text-to-text","license":"mit","downloads":3648,"architecture":"deepseek_v41","dtype":"NVFP4","createdAt":"2026-10-30T17:37:02.000Z","published":"上线 2026-10-31","releasedAt":"2026-10-30T17:37:02.000Z"},{"id":"hf-sensenova-loop-9b","time":"11-05 12:50","slot":"digest","tag":"多模态","title":"sensenova/SenseNova-Loop-9B-Preview","why":"Loop 多模态预览；looped-transformer / MHC 架构标签","params":"≈9.7B · sensenova_loop · ctx 256K · Apache-2.0 · 下载 2.7K · image-text-to-text","source":"Hugging Face","href":"https://huggingface.co/sensenova/SenseNova-Loop-9B-Preview","accent":2,"pipeline_tag":"image-text-to-text","license":"apache-2.0","downloads":2704,"architecture":"sensenova_loop","context_length":262144,"createdAt":"2026-10-28T15:34:40.000Z","published":"上线 2026-10-29","releasedAt":"2026-10-29T13:46:34.000Z"},{"id":"hf-qwen35-4b-3d-coder","time":"11-05 12:50","slot":"digest","tag":"LLM","title":"ccore/Qwen3.5-4B-3D-Coder","why":"基于 Qwen3.5-4B 的 3D/Three.js 代码微调；含 GGUF","params":"≈4.5B · qwen3_5 · F16 · ctx 256K（GGUF）· Apache-2.0 · 下载 123K","source":"Hugging Face","href":"https://huggingface.co/ccore/Qwen3.5-4B-3D-Coder","accent":1,"pipeline_tag":null,"license":"apache-2.0","downloads":123321,"architecture":"qwen3_5","dtype":"F16","context_length":262144,"createdAt":"2026-10-25T05:43:34.000Z","published":"上线 2026-10-25","releasedAt":"2026-10-25T05:43:34.000Z"},{"id":"hf-nanbeige43-33b","time":"11-05 12:50","slot":"digest","tag":"LLM","title":"Nanbeige/Nanbeige4.3-33B","why":"中文向 33B 对话模型，llama 架构，TGI/endpoints 兼容","params":"33.1B · llama · BF16 · ctx 512K · Apache-2.0 · 下载 18.7K · text-generation · TGI","source":"Hugging Face","href":"https://huggingface.co/Nanbeige/Nanbeige4.3-33B","accent":0,"pipeline_tag":"text-generation","license":"apache-2.0","downloads":18657,"architecture":"llama","dtype":"BF16","context_length":524288,"createdAt":"2026-10-24T02:49:18.000Z","published":"上线 2026-10-24","releasedAt":"2026-10-24T11:57:14.000Z"},{"id":"hf-lfm3-1.7b","time":"11-05 12:50","slot":"digest","tag":"边缘","title":"LiquidAI/LFM3-1.7B","why":"边缘小模型趋势前列；另有 LFM3-1.7B-GGUF（llama.cpp，ctx 256K）","params":"1.7B · lfm3 · BF16 · ctx 256K · license:other · 下载 55K · text-generation","source":"Hugging Face","href":"https://huggingface.co/LiquidAI/LFM3-1.7B","accent":1,"pipeline_tag":"text-generation","license":"other","downloads":54959,"architecture":"lfm3","dtype":"BF16","context_length":262144,"createdAt":"2026-10-23T21:35:42.000Z","published":"上线 2026-10-24","releasedAt":"2026-10-23T21:35:42.000Z"},{"id":"hf-deepseek-v41-flash-1017","time":"11-05 12:50","slot":"digest","tag":"多模态","title":"deepseek-ai/DeepSeek-V4.1-Flash-1017","why":"全站与 text-generation 榜双高，FP8/多模态服务端热度","params":"≈763B · deepseek_v41 · FP8/8-bit · ctx 1M · MIT · 下载 222K · image-text-to-text · endpoints","source":"Hugging Face","href":"https://huggingface.co/deepseek-ai/DeepSeek-V4.1-Flash-1017","accent":0,"pipeline_tag":"image-text-to-text","license":"mit","downloads":221621,"architecture":"deepseek_v41","dtype":"FP8","context_length":1048576,"createdAt":"2026-10-17T02:40:38.000Z","published":"上线 2026-10-17","releasedAt":"2026-10-17T02:40:38.000Z"},{"id":"hf-nemotron-transcribe-30b","time":"11-05 12:50","slot":"digest","tag":"ASR","title":"nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B","why":"非 LLM：ASR/转写旗舰（30B-A4B），多语种长音频与说话人分离标签","params":"≈33.7B · nemo_speechlm · license:other · 下载 45K · automatic-speech-recognition","source":"Hugging Face","href":"https://huggingface.co/nvidia/NVIDIA-Nemotron-3.5-Transcribe-30B-A4B","accent":0,"pipeline_tag":"automatic-speech-recognition","license":"other","downloads":44939,"architecture":"nemo_speechlm","createdAt":"2026-10-07T22:50:21.000Z","published":"上线 2026-10-08","releasedAt":"2026-10-07T22:50:21.000Z"},{"id":"hf-mistral-large4-1t-a52b","time":"11-05 12:50","slot":"digest","tag":"MoE","title":"mistralai/Mistral-Large-4-1T-A52B","why":"1T/A52B MoE 旗舰多模态；vLLM / SGLang / llama.cpp 标签齐全","params":"型号 1T/A52B MoE · mistral_large4 · FP8 · ctx 1M · license:other · 下载 8.2K · vLLM/SGLang/llama.cpp","source":"Hugging Face","href":"https://huggingface.co/mistralai/Mistral-Large-4-1T-A52B","accent":0,"pipeline_tag":"image-text-to-text","license":"other","downloads":8200,"architecture":"mistral_large4","dtype":"FP8","context_length":1048576,"serving":["vLLM","SGLang","llama.cpp"],"createdAt":"2026-09-30T18:35:13.000Z","published":"上线 2026-10-01","releasedAt":"2026-09-30T18:35:13.000Z"},{"id":"hf-qwen38-27b","time":"11-05 12:50","slot":"digest","tag":"多模态","title":"Qwen/Qwen3.8-27B","why":"旗舰多模态；旁路 nvidia/Qwen3.8-Flash-Next-NVFP4 下载量大","params":"27.8B · qwen3_5 · ctx 256K · Apache-2.0 · 下载 8.0M+ · image-text-to-text · endpoints","source":"Hugging Face","href":"https://huggingface.co/Qwen/Qwen3.8-27B","accent":2,"pipeline_tag":"image-text-to-text","license":"apache-2.0","downloads":8032442,"architecture":"qwen3_5","context_length":262144,"createdAt":"2026-08-05T08:22:59.000Z","published":"上线 2026-08-14","releasedAt":"2026-08-14T15:00:00.000Z"}]}