Nvidia Releases Open-Weight Nemotron 3.5 Transcribe Model for Multilingual Voice AI
aiinsider ·
Nvidia has released Nemotron 3.5 Transcribe, a free open-weight model designed for multilingual voice AI applications. The model supports transcription in up to 40 languages and can translate between 31 languages and English. It is capable of processing recordings lasting up to an hour while distinguishing between eight different speakers. Nvidia claims the model outperforms competitors on benchmarks such as Hugging Face’s Open ASR Leaderboard. With 30 billion parameters, it can run on single chips ranging from edge devices like Jetson AGX Thor to data center processors. The release targets developers building tools for business meetings and medical transcription using industry-specific terminology. This launch adds to Nvidia's portfolio of open-weight voice models, including the smaller Parakeet variant. The announcement comes amid a competitive period for voice AI featuring releases from Mistral, ElevenLabs, and Salesforce.
aiinsider