The distilled version
- Nvidia released Nemotron 3.5 Transcribe, a free, open-weight model that transcribes recorded audio in 40 languages and tells speakers apart.
- It’s aimed at developers building meeting and medical transcription apps, so you won’t use it directly, but it could lower costs in tools built on it.
- Watch for your transcription or meeting-notes tool to mention it, then compare pricing and accuracy before switching.
What happened
Nvidia released Nemotron 3.5 Transcribe on October 28, a free, open-weight model that transcribes recorded audio in up to 40 languages, translates 31 of them to and from English, and can tell up to eight speakers apart in a recording up to an hour long. Nvidia also released version 2 of Granary, its open speech dataset used to train and adapt the model.
According to Nvidia, the model leads among open-weight models on several outside benchmarks, including Hugging Face’s Open ASR Leaderboard. It is built for developers, available now on Hugging Face and ModelScope, and can run on anything from a single chip to Nvidia’s larger datacenter hardware. It works only with recorded audio, not live streams.
The release lands amid a wave of voice AI launches, including Mistral’s Voxtral 3, ElevenLabs’ Realtime, and Salesforce’s Launa-Omni, according to TechCrunch.
Why it matters
If your business already pays for a transcription or meeting-notes tool, this doesn’t change your software today. But because the model is free and open, companies that build those tools can use it to cut their own costs, which could eventually mean cheaper plans or better accuracy for the transcription and meeting-summary features you already use.
What it means for your business
If you run the business: there’s nothing to install here. The model targets developers, not end users. Keep an eye on whether the meeting or call-transcription tools you already pay for mention using it, since that could affect pricing or quality.
What to do
Worth watching- Check whether your meeting-notes or call-transcription tool shares which AI model it uses.
- If you’re evaluating transcription vendors, ask whether they plan to adopt newer open models to lower their costs.
- No direct setup needed this week since the model requires developer tools to run.
Worth watching: no action needed yet; revisit when it matures.
Jargon in this story
Sources
- 1Voice AI in vogue: Nvidia releases new open transcription model (opens in a new tab) · TechCrunch · News report · Oct 28, 2026
- 2NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt (opens in a new tab) · Hugging Face · Company blog · Oct 28, 2026