Nvidia Releases Open-Weight Nemotron 3.5 Transcribe Model for Multilingual Voice AI
By James Dargan · · 1 min read
Nvidia has released Nemotron 3.5 Transcribe, a free, open-weight model that transcribes audio in up to 40 languages, alongside version 2 of Granary, its open source speech dataset.
The company said the model can translate 31 languages to and from English, process recordings of up to an hour, and distinguish between as many as eight speakers. Nvidia claims it outperforms rivals on benchmarks including Hugging Face’s Open ASR Leaderboard.
The tools are aimed at developers building applications for business meetings and medical transcription, two of AI’s most established workplace uses. Nvidia said developers can combine the model, the updated Granary dataset and tools such as NeMo Speech to tailor apps to recognize industry-specific terminology in each language. However, the model only handles recorded audio, not live streams.
At 30 billion parameters, Transcribe can run on a single chip, ranging from the Jetson AGX Thor robotics controller and DGX Spark chips in new laptops to Blackwell data center processors. It adds to Nvidia’s lineup of open-weight voice models, including the compact 1.1-billion-parameter Parakeet.
The release arrives during a busy stretch for voice AI, following Mistral’s Voxtral 3, ElevenLabs’ Realtime for instant speech in voice agents, and Salesforce’s Launa-Omni voice model.
