Nvidia releases Nemotron 3.5 Transcribe, a free open-weight speech model for 40 languages
The 30-billion-parameter model comes with version 2 of Nvidia's Granary speech dataset and can run on a single chip.
Nvidia released Nemotron 3.5 Transcribe on Wednesday. It is a free, open-weight model that transcribes audio in up to 40 languages, and Nvidia also released version 2 of Granary, its open-source speech dataset. The model is listed on Hugging Face as NVIDIA-Nemotron-3.5-Transcribe-30B-A4B. Nvidia says it can translate 31 languages to and from English, handle recordings up to an hour long, and tell up to eight speakers apart. The company claims it beats rivals on benchmarks, including Hugging Face's Open ASR Leaderboard. The model has 30 billion parameters, and Nvidia says it runs on a single chip, from the Jetson AGX Thor robotics controller and DGX Spark to Blackwell datacenter chips. It works only on recorded audio, not livestreams. Nvidia is aiming at developers who build apps for business meetings and medical transcription, and it says the updated dataset and tools such as NeMo let teams adapt the model to an industry's jargon in each language. The model joins Nvidia's smaller open voice models, such as the 1.1-billion-parameter Parakeet. It arrives amid a busy period for voice AI that includes Mistral's Voxtral 3. Open weights let organizations transcribe sensitive audio on their own hardware instead of sending it to a cloud service.