Most transcription pipelines are a relay race: one model detects speech, another transcribes, another figures out who said what, and something else answers questions. Nvidia wants to replace the whole team with one runner.
On Wednesday, Nvidia released Nemotron 3.5 Transcribe, an open-weight speech model that handles transcription, speaker attribution, translation and questions about the audio in a single model pass.
What it can do
According to SiliconANGLE, the model can work through 100 hours of audio, covers 40 languages and can label up to eight speakers. It is a 30-billion-parameter mixture-of-experts model with about 4 billion active parameters, so it fits on fairly modest hardware.
It’s promptable, too. Nvidia’s examples include “Transcribe this recording and label each speaker,” “Translate this audio into English” and “List what decisions were made in this meeting.” Users can feed in product terms and jargon ahead of time, which Nvidia says helps in fields like healthcare, financial services and customer service.
Performance and hardware
Nvidia says the model ranks No. 1 on the Hugging Face Open ASR leaderboard and No. 2 among open-weight models on Artificial Analysis. It runs on a single Nvidia GPU, from Jetson AGX Thor and DGX Spark up to Blackwell data center GPUs, and Nvidia says a Blackwell Server Edition GPU using NVFP4 precision processes 100 hours of audio in under three minutes.
Nvidia also expanded its open Granary speech dataset to version 2, which covers 463,062 hours across 48 languages, and released an open data pipeline for fine-tuning. The model is available now on Hugging Face, with quantized versions coming for local use on RTX Spark laptops and other systems.
Why it matters
Because the weights are open, companies can choose where their audio is processed, from a laptop to their own cloud. That matters for regulated industries and for contact centers that handle sensitive customer calls.
For UCaaS, CCaaS and conversation intelligence vendors, the bar for built-in transcription and call analytics just went up. Buyers will expect meeting summaries, multilingual support and decision tracking as standard, and vendors that clearly explain their AI and data-handling story will have an easier time turning interest into pipeline.
Sources: SiliconANGLE, Nvidia. Written by the xRapid Media LLP team.
Selling to the IT market? xRapid Media LLP is a High Tech B2B Marketing company. We help technology companies reach decision-makers with verified B2B data, account-based campaigns, email and LinkedIn outreach and appointment setting. Call +91 9767196309 or book a free B2B marketing consultation.