Nvidia releases Nemotron 3.5 Transcribe speech model as open weights

Nvidia released Nemotron 3.5 Transcribe, a 30-billion-parameter open-weight speech model, on Hugging Face, according to TechCrunch. The company also released version 2 of Granary, its open-source speech dataset.

The model transcribes audio in up to 40 languages and translates 31 languages to and from English, TechCrunch reported. It handles recordings up to one hour long and distinguishes up to eight speakers. It works on recorded audio only, not livestreams.

Nvidia said the model can run on a single chip. The supported hardware ranges from Jetson AGX Thor and DGX Spark to Blackwell datacenter chips. The company said the model beats rivals on benchmarks, including Hugging Face's Open ASR Leaderboard. The article does not give specific scores.

The article describes the model as "free, open-weight". It does not name a specific licence. No named speakers are quoted in the article.

Nvidia also offers Parakeet, a multilingual open-weight voice model with 1.1 billion parameters. The article does not say how Nemotron 3.5 Transcribe compares with Parakeet.

TechCrunch's report, Voice AI in vogue: Nvidia releases new open transcription model, is the primary source for these details. The article does not say when the licence terms will be published.