NVIDIA’s reported transcription-model release adds benchmark and size details, but the reported score differs from the earlier recorded English leaderboard figure and requires clarification.
Limits of the evidence: The release and performance figures come from NVIDIA’s official blog; the supplied text does not establish independent reproduction or specify the weights’ licence.
What to watch
Assessment revised Oct 29, 2026
NVIDIA Nemotron 3.5 Transcribe and Granary v2: Intelligent Multilingual Transcription You Can Prompt and Adapt
Hugging Face BlogOct 28, 2026
Voice AI in vogue: Nvidia releases new open transcription model
TechCrunch AIOct 28, 2026First report
NVIDIA Releases Nemotron 3.5 Transcribe: An Open 30B MoE Model for 40-Language Transcription with ~4B Active Parameters
MarkTechPostOct 29, 2026
Contradicting report from MarkTechPost: Nemotron 3.5 Transcribe scored 4.71% mean WER on the English Open ASR Leaderboard.
Older report added; assessment kept from a newer one
First assessment, from TechCrunch AI
Oct 29, 2026
Earlier claim
Nemotron 3.5 Transcribe had an English average WER of 5.33% on the Hugging Face Open ASR Leaderboard.
Later claim
Nemotron 3.5 Transcribe scored 4.71% mean WER on the English Open ASR Leaderboard.
No maturity ladder applies to this desk.
Sens.ai aggregates and assesses published reporting. The assessment above is machine generated; the original sources are authoritative.