NVIDIA Nemotron 3.5 Transcribe: open-weight speech model that transcribes, translates and attributes speakers in one pass
Open-weight 30B MoE speech model (about 4B parameters active per token) covering 40 languages, released with Granary v2, an open speech dataset and curation pipeline spanning 463,062 hours across 48 languages.
Why it matters
One promptable checkpoint replaces a chained ASR plus diarization plus translation pipeline, runs on a single GPU, and tops open-weight ASR leaderboards.
What you could build with it
A small transcription agency could offer speaker-labelled, translated meeting minutes for multilingual clients by fine-tuning this model on each client's domain vocabulary with the Granary v2 pipeline, charging per recorded hour instead of building a speech stack from scratch.
Does it hold up?
Usable today for recorded-audio transcription via the Hugging Face demo and hosted endpoints; NVIDIA's medical case study cut word error rate 45.8% relative after fine-tuning, but streaming is explicitly unsupported, so live-caption use cases should look elsewhere.
Built with NVIDIA Nemotron 3.5 Transcribe
- Voice AI in vogue: Nvidia releases new open transcription modeltechcrunch · Reports the Oct 28 release, the 40-language coverage, the single-GPU claim, and the recorded-audio-only limitation.
- Nvidia releases all-in-one adaptable Nemotron 3.5 audio transcription and translation modelsiliconangle · Covers prompt-steered outputs, Granary v2's three transcript styles, and the Hugging Face demo.
- Nvidia Releases Open Nemotron 3.5 Transcribe Speech Modelaiparabellum · Walks through the one-checkpoint design, jargon prompting, and the single-GPU deployment range.
Learn more
First spotted on article: source.
More AI models
OpenAI publishes 722 AI-generated math manuscripts from an unreleased modelOpenAI released 722 mathematical manuscripts in 372 result families to GitHub under Apache-2.0, produced by an…models · JEV 0.74Tencent Hunyuan Hy3: open 295B MoE with 21B active under Apache 2.0Tencent open-sourced Hy3, a 295B-parameter mixture-of-experts model with only 21B parameters active per token, under…models · JEV 0.73NVIDIA Isaac Lab 3.0 GA + Isaac OS open runtime + Agile ONE SN humanoidMajor Isaac platform expansion at GTC Berlin Oct 21: Isaac Lab 3.0 is generally available as the open-source robot…models · JEV 0.73StepFun Step-5-Preview — open BF16 weights releasedStepFun published the full BF16 weights for its flagship Step-5-Preview agentic model (600B-parameter sparse MoE, 27B…models · JEV 0.73NVIDIA Nemotron 3.5 Retriever collection: Embed 8B/1B, Rerank 8B, Parse 631MNVIDIA's Oct 21 collection drop for agentic retrieval: Nemotron-3.5-Embed in 8B BF16, 1B BF16 and 1B NVFP4…models · JEV 0.72Kimi K3.1 Pro: Moonshot's 2.8T open-weight flagship officially launchesMoonshot AI officially launched Kimi K3.1 Pro on Oct 23, a mixture-of-experts model with 2.8 trillion total / 104…models · JEV 0.72
Get the week's best AI launches, plus 3 ideas worth building
One email every Saturday. Ranked by traction, not hype. Free.