[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-slug-nemotron-3-5-asr-nvidia-600m-40-lang":3,"topics-all":36,"news-related-ab856198-bde1-4c2a-9ca2-77ccecf97cbd":55},{"id":4,"title":5,"summary":6,"content":6,"original_url":7,"source_id":8,"tags":9,"translations":23,"news_slug":29,"published_at":30,"created_at":31,"modified_at":32,"is_published":33,"publish_type":34,"image_url":13,"view_count":35},"ab856198-bde1-4c2a-9ca2-77ccecf97cbd","ASR流式新标杆：NVIDIA Nemotron 3.5 ASR以600M参数覆盖40语种","NVIDIA 在 6 月 4 日把 Nemotron 3.5 ASR 摆上了 Hugging Face：600M 参数，一个 checkpoint，覆盖 40 种语言-地区；流式输出内置大小写与标点，许可证为 OpenMDW-1.1。这不是又一次「Whisper 后继者」式复刻，而是把语音 AI 的「流式 vs 准确率」长期死结拆掉的工程样本。\n\n核心是 Cache-Aware FastConformer-RNNT。传统流式 ASR 用滑窗反复重算重叠音频块，把一半算力浪费在已经处理过的 token 上；Nemotron 3.5 改成缓存 encoder 的自注意力和卷积激活，每一帧只算一次，不重复劳动。NVIDIA 自报在 H100 上比基线提高 17 倍并发流数。更妙的是 att_context_size 这一参数化旋钮：80ms、160ms、320ms、560ms、1.12s 五档延迟，对应同一组权重，部署时再选择，无需重训。\n\n最容易被忽视的是「单 checkpoint 多语种」的工程意义。常见做法是给每种语言部署一个独立模型，再叠一个语言识别组件；这里直接用 prompt-based target_lang 条件输入做语言路由，target_lang=auto 模式还能在终末标点后自动输出语种标签。客服、会议转写、多语种混合播客等场景再也不需要 40 个模型的「博物馆」。\n\n对照商业 API：Whisper large-v3 跑离线批处理，Deepgram Nova-3、AssemblyAI Universal-3 Pro、ElevenLabs Scribe v2 Realtime 都是闭源流式，但不支持本地化微调。Nemotron 3.5 开放权重意味着可针对希腊语、保加利亚语等小语种做继续训练——NVIDIA 给出的样本里，希腊语 FLEURS WER 从 35 降到 24（相对 32%），保加利亚语从 22 降到 15（相对 31%），都是在最激进的 80ms 模式下跑出来的。\n\n语音智能体（voice agent）爆发前夜，吞吐和首字延迟是生产级分水岭，Nemotron 3.5 至少把「自托管、低延迟、可微调」这三条首次压进了一个 600M 模型里。真正的考验是 NIM 上的 gRPC 流式接口何时放出——开源和商业部署之间的最后一公里，往往决定一个模型到底能不能进生产。","https:\u002F\u002Fhuggingface.co\u002Fblog\u002Fnvidia\u002Ffine-tuning-nemotron-35-asr","474eef8c-e0c3-46cf-adee-c089558220f9",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"fca9258a-9430-455a-b95d-b9fae5e373a8","ai-inference",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"7e89b5cc-57db-4f37-bc6d-28919a73931c","model-release",{"id":18,"name":19,"slug":19,"description":13,"color":13},"8dac812d-3839-4abe-a855-5f56ec9515fd","nvidia",{"id":21,"name":22,"slug":22,"description":13,"color":13},"b9bd9039-fcdb-41a8-b85b-fc1587def2b9","open-source",[24],{"id":25,"lang":26,"title":27,"summary":28,"content":13},"2d5a2c8e-1470-421e-99ab-e9f00fd4e398","en","Nemotron 3.5 ASR: streaming benchmark, 40 languages at 600M","NVIDIA put Nemotron 3.5 ASR on Hugging Face on June 4: 600M parameters, one checkpoint, 40 language-region coverage; streaming output has built-in casing and punctuation, licensed under OpenMDW-1.1. This isn't another \"Whisper successor\" remake — it's an engineering sample of breaking the long-standing \"streaming vs. accuracy\" deadlock in speech AI.\n\nThe core is Cache-Aware FastConformer-RNNT. Traditional streaming ASR uses sliding windows to repeatedly recompute overlapping audio chunks, wasting half the compute on tokens already processed; Nemotron 3.5 instead caches the encoder's self-attention and convolution activations, computing each frame only once without repeated work. NVIDIA self-reports a 17× increase in concurrent streams on H100 over the baseline. The nicest part is the `att_context_size` parameterized knob: five latency tiers (80ms, 160ms, 320ms, 560ms, 1.12s) share the same weights, with the choice made at deployment time without retraining.\n\nThe most easily overlooked is the engineering meaning of \"single checkpoint, many languages.\" The common practice is to deploy a separate model per language and stack a language-identification component on top; here the system uses a prompt-based `target_lang` conditional input for language routing, and `target_lang=auto` can even output a language tag after terminal punctuation. Customer service, meeting transcription, multilingual mixed podcasts no longer need a \"museum\" of 40 models.\n\nCompared with commercial APIs: Whisper large-v3 runs offline batch processing, Deepgram Nova-3, AssemblyAI Universal-3 Pro, and ElevenLabs Scribe v2 Realtime are all closed-source streaming models that don't support local fine-tuning. Nemotron 3.5's open weights mean continued training is possible for small languages like Greek and Bulgarian — the samples NVIDIA provides show Greek FLEURS WER dropping from 35 to 24 (32% relative), and Bulgarian from 22 to 15 (31% relative), both under the most aggressive 80ms mode.\n\nOn the eve of the voice-agent explosion, throughput and time-to-first-token are the production-grade dividing line, and Nemotron 3.5 at least stuffs \"self-hosted, low-latency, fine-tunable\" into a 600M model for the first time. The real test is when gRPC streaming interfaces on NIM ship — the last mile between open source and commercial deployment often decides whether a model can actually go into production.","nemotron-3-5-asr-nvidia-600m-40-lang","2026-06-11T06:30:00Z","2026-06-11T06:28:53.173102Z","2026-08-19T02:08:40.142862Z",true,"agent",177,[37,46],{"slug":38,"tag_slug":38,"title_zh":39,"title_en":40,"intro_zh":41,"intro_en":42,"id":43,"is_active":33,"created_at":44,"modified_at":45},"ai-for-science","AI for Science 2026：从 UniPert 到 GPT-Rosalind 的硬核进化","AI for Science 2026: from UniPert to GPT-Rosalind","生命科学、化学材料、物理世界模型——AI 正在从\"语言工具\"变成\"实验伙伴\"。本专题收录 AI 在三大科学方向的关键节点：UniPert 统一基因与化学扰动空间、GPT-Rosalind 端到端生命科学推理、达摩院 AI 智能体 28 小时找到 4 种超导新材料、Anthropic Claude Science 把工作台做成标准品。","From language tool to lab partner — AI is reshaping life sciences, chemistry\u002Fmaterials, and physical world models. This topic covers the key milestones: UniPert unifying genetic-chemical perturbation spaces, GPT-Rosalind's end-to-end life-sciences reasoning, DAMO's AI agent discovering 4 superconducting materials in 28 hours, and Anthropic's Claude Science workbench going mainstream.","988a4300-5fab-41c4-b5d8-63711a2dc757","2026-09-10T01:34:15.296649Z","2026-09-10T01:34:15.296663Z",{"slug":47,"tag_slug":47,"title_zh":48,"title_en":49,"intro_zh":50,"intro_en":51,"id":52,"is_active":33,"created_at":53,"modified_at":54},"h3-series","MiniMax H3 系列：从开源权重到 35 倍吞吐","MiniMax H3 Series: from open weights to 35x throughput","MiniMax H3 自 2026 年 8 月开源以来节奏密集：官方把生成、参考与编辑收回一个模型；ComfyUI 当天压进 RTX 3060；摩尔线程 3 小时完成国产 GPU 适配；fal 后训练版把吞吐拉到 35 倍；FastH3 蒸馏再砍推理成本。本专题持续追踪 H3 的发布—开源—蒸馏—部署全链路。","Since MiniMax open-sourced H3 in August 2026 the pace has been relentless: one unified omni-modal model, same-day ComfyUI support down to an RTX 3060, a 3-hour Day-0 port to Moore Threads GPUs, fal's post-trained H3 Max at 35x throughput, and FastH3 distillation cutting inference cost further. This topic tracks the full H3 chain — release, open weights, distillation, deployment.","83ef0daa-3c31-4cb3-86ed-e5ee58654d5f","2026-09-08T07:33:19.942193Z","2026-09-08T07:33:19.942209Z",{"items":56},[57,62,67,72,77,82],{"id":58,"title":59,"news_slug":60,"published_at":61},"61cf8d85-c751-4da2-9aae-10b645415ec9","英伟达发布开源工具 PAIR,把家里电脑连成个人 AI 推理集群","nvidia-pair-personal-ai-router-local-inference","2026-09-09T02:00:00+00:00",{"id":63,"title":64,"news_slug":65,"published_at":66},"107277b2-2c3f-490b-bb68-a1432e723649","英伟达开源 PAIR：把家里闲置显卡串成一座个人 AI 数据中心","nvidia-pair-personal-ai-router","2026-09-05T06:25:00+00:00",{"id":68,"title":69,"news_slug":70,"published_at":71},"42b7939c-1b44-43b8-95cf-a8fc2204560d","NVIDIA 开源 Personal AI Router，把家里 RTX 与 Mac 拼成本地 AI 集群","nvidia-personal-ai-router-pair-beta","2026-09-04T03:20:00+00:00",{"id":73,"title":74,"news_slug":75,"published_at":76},"61de017b-bdd6-44b3-9f45-d4fb233bd24d","PhoneLLM 开源:30B MoE 电话客服模型,自称比 GPT-5.6 Terra 便宜 94%","phonellm-alpha-1-voice-agent-open-model","2026-08-29T21:10:00+00:00",{"id":78,"title":79,"news_slug":80,"published_at":81},"de2cceb2-7d39-4a5f-844e-5a3144667f49","Nemotron 3.5 Lightning 开源：30B 总参 3B 激活的混合 MoE，直接用 NVFP4 配方预训练","nemotron-35-lightning-30b-a3b-open-release","2026-08-16T15:00:00+00:00",{"id":83,"title":84,"news_slug":85,"published_at":86},"bcdc10bc-2f08-4c39-8ffa-e7e34041c112","京东开源 JoyAI-Video-Edit:用 16B 多模态扩散 Transformer 把视频编辑推进「边播边改」实时流时代","jd-joyai-video-edit-real-time-streaming","2026-08-05T03:00:00+00:00"]