NVIDIA launches open-weight 30B MoE transcription model Nemotron 3.5 Transcribe
baseten · x · 2026-10-28
NVIDIA released Nemotron 3.5 Transcribe, an open-weight speech-understanding model: a 30B MoE with 4B active parameters that transcribes in 40 languages.
It unifies multilingual ASR (including code-switching), speaker attribution, speech-to-text translation, contextual prompting (inject product names and domain terminology to guide recognition), and audio Q&A into one checkpoint — targeting recorded-audio workflows like post-call analytics, meeting tools, media transcription, and voice-of-customer analysis. Licensed under OpenMDW-1.1. Baseten announced it is a day-0 launch partner for deployment.
More from Multimodal
- Dev vibecodes a ComfyUI node that auto-archives and organizes all your generations — LaurentLaSalle · 2026-10-29
- Civitai models can legally be shared for free, Buzz paywalls don't restrict redistribution — ArtyfacialIntelagent · 2026-10-29
- Same Seedance 2.5 video costs $2.32 on Sunra vs $4.62 on Higgsfield — hey_abusiddik · 2026-10-29
- Higgsfield launches Selected showcase; first pick MADSEASON is fully open source — CurieuxExplorer · 2026-10-29
- Wistia launches Mochine, an AI video editor that adds motion graphics to talking-head videos — Scobleizer · 2026-10-29
- Fizgig H3 still-image latent node drops into any ComfyUI workflow — and runs faster — robertwellesley · 2026-10-29