Nvidia released Nemotron 3.5 Transcribe on Wednesday, a free, open-weight model that transcribes audio in up to 40 languages. Why it matters: a 30-billion-parameter transcription model that can run on a single chip gives developers an open option for meeting and medical transcription instead of a closed API.
October 28, 2026
What happened?
Nvidia released Nemotron 3.5 Transcribe and version 2 of Granary, its open-source speech dataset. According to Nvidia, the model can translate 31 languages to and from English, handle recordings up to an hour long, and tell up to eight speakers apart.
Nvidia says the model beats rivals on benchmarks, including Hugging Face’s Open ASR Leaderboard. Those are company-reported results, not independent tests.
What are the details?
The tools are aimed at developers building apps for business meetings and medical transcription. Nvidia says developers can combine the model, the updated Granary dataset, and tools such as its NeMo Speech model to customize an app to recognize an industry’s jargon in each language.
At 30 billion parameters, Nvidia says Transcribe can run on a single chip, from the Jetson AGX Thor robotics controller and the DGX Spark chip powering new laptops to Blackwell datacenter chips. It joins Nvidia’s other open-weight voice models, including Parakeet at 1.1 billion parameters.
One clear limit: Transcribe works with recorded audio, not livestreams.
Why it matters
Voice is having a busy month. Mistral shipped Voxtral 3 earlier this month, ElevenLabs launched Realtime for instant speech in voice agents, and Salesforce released Launa-Omni, its voice model. Nvidia’s entry is different in one practical way — it is open-weight, so teams can run and customize it themselves rather than sending every recording to someone else’s cloud.
For Tested Intelligence readers, the test is not the leaderboard claim. It is whether a single-chip, open model that handles long recordings, multiple speakers, and dozens of languages makes transcription cheap and private enough to build into everyday tools. That is everything AI, tested — the claim is out, the real-world checks come next.
Frequently asked questions
Is Nemotron 3.5 Transcribe free?
Yes. TechCrunch describes it as a free, open-weight model, released alongside version 2 of the open-source Granary speech dataset.
How many languages does it support?
Nvidia says it transcribes audio in up to 40 languages and can translate 31 languages to and from English.
Can it transcribe live audio?
No. According to the release coverage, Transcribe works with recorded audio, not livestreams. It handles recordings up to an hour long and can distinguish up to eight speakers.
Sources: TechCrunch