NVIDIA introduces Nemotron 3.5 Transcribe with transcription in 40 languages
NVIDIA describes speaker attribution and translation in one checkpoint; SiliconANGLE reports a demo and downloadable checkpoint.
TLDR
NVIDIA introduced Nemotron 3.5 Transcribe, an open-weight model it says handles transcription in 40 languages alongside speaker attribution, translation and audio understanding. SiliconANGLE reports that developers can try a demo and download a checkpoint for fine-tuning. Artificial Analysis reports that the model tied for first among open-weight models on its AA-WER Index, a result specific to that benchmark.
Combined views
1.1M
25 Sources, first seen ago

Comments