Nvidia releases open Nemotron 3.5 Transcribe speech model
Nvidia has released Nemotron 3.5 Transcribe, an open-weights speech model that turns audio into text. It handles 40 languages and adds speaker labels and word timestamps. It reports a 4.71% word error rate on the English Open ASR Leaderboard, a public test ranking.
Quick glossaryNvidia, open-weights, word error rate
- Nvidia
- A US chipmaker best known for the graphics processors that power most AI training.
- open-weights
- A model whose trained settings are free for anyone to download and run on their own computers.
- word error rate
- A score showing how often a speech tool gets words wrong; lower numbers mean better accuracy.
Burning Potato score: 65%