// AI news · 2026-10-28 · Hugging Face Blog
NVIDIA Releases Nemotron 3.5 Transcribe for Multilingual Transcription
NVIDIA releases Nemotron 3.5 Transcribe, an open-weight multilingual transcription model supporting 40 languages, speaker diarization in 13, and translation across 31 languages.
Published 2026-10-28 22:58 (Beijing time)Source Hugging Face BlogAuto-summarised, fact-checked against source
01Summary
It ranks first on Hugging Face Open ASR Leaderboard with 5.33% WER in English and leads in multilingual and long-form tasks. The model supports prompt-based control, local deployment on devices like RTX Spark or Jetson AGX Thor, and domain-specific fine-tuning. In a medical case study, fine-tuning reduced WER by 45.8% relative to baseline.
Summarised from the original article by an open-source model, then checked automatically: every number and English name must appear in the source. The original article is authoritative.
02Key points
- Supports 40 languages for transcription, 13 for speaker diarization, 31 for English translation.
- Ranks first on Hugging Face Open ASR Leaderboard with 5.33% WER in English.
- Offers promptable outputs and fine-tuning via Granary v2 pipeline for domain adaptation.
- Reduces WER by 45.8% in medical domain after fine-tuning.
03Source
Hugging Face Blog · published 2026-10-28 22:58 Beijing time
Also reported by 3 other outlets
05Related news
- AWS Enables Ray Train DLC Distributed Training on EKS2026-10-29 · AWS Machine Learning Blog
- NVIDIA RTX 5090 Faces Potential Discontinuation2026-10-27 · IT之家
- SWI Compute Plans 530MW Nvidia AI Campus in UK2026-10-27 · The Next Web
- NVIDIA Releases Global Performance Tuner for TensorRT2026-10-24 · NVIDIA
- Jensen Huang's GTC Berlin 2026 Keynote2026-10-21 · NVIDIA Blog
- Samsung Charges Up to 25% Premium for HBM Long-Term Supply2026-10-25 · cnBeta
From AIMap AI news: collected from 1,100+ official and media sources. Summaries only restate the original article; follow the link for the full story. All AI news · Today