Nemotron 3.5 ASR Streaming 0.6b
NVIDIA · released May 15, 2026 · nvidia/nemotron-3.5-asr-streaming-0.6b
Speech to textTranscribes a recording into words
- Type
- Open weightsCustom licence
- Languages
- 35
- Size
- 0.6B
Context measured in tokens · download allowed, licence restricts use
Our take
Written Sep 4, 2026Nemotron 3.5 ASR Streaming is a tiny downloadable speech-to-text model from NVIDIA that turns audio into written words across 35 languages. It is built for speed rather than accuracy, transcribing an hour of audio in about two seconds on benchmark hardware.
Pick this when raw speed is non-negotiable — real-time streaming pipelines or low-latency applications where throughput beats precision. Use it for multilingual coverage from a single small model, or for self-hosted financial earnings calls where it performs best. Skip it if you need accurate meeting transcription, heavily accented speech, or podcast-quality output, where its error rate rises sharply.
The case for it
- Transcribes at roughly 1,490 times real time — an hour of audio in about two seconds.
- 35 languages including major European, Asian and Middle Eastern languages from a 0.6-billion-parameter model.
- OpenMDW 1.1 licence and small size make it easy to self-host on modest hardware.
- Strongest on formal financial audio, with about one word in thirty wrong on earnings calls.
The case against it
- Below-median accuracy on every condition we track, from clean read-aloud to meetings and accented speech.
- Error rate more than quadruples on meeting audio and quintuples on accented speech versus its clean-read score.
- No commercial hosts listed, so you must run it yourself.
How good is it?
A small open speech-to-text model for turning recordings into text, though it trails most models on accuracy.
- turning spoken English into written textOpen ASR WER · 67th of 76
- transcribing recordings of meetings in a roomRecorded meetings · 70th of 92
- transcribing speakers with a range of accentsAccented speech · 69th of 76
- transcribing podcasts and video audioPodcasts and video · 77th of 92
- transcribing clear recordings of people reading aloudClean read speech · 82nd of 92
TranscriptionTurning speech into text1.5 of 5Open ASR WER · 67th of 76
92.1%
Misses roughly one word in 13, averaged over nine English test sets.
1,345×30th of 74
an hour of audio in 3 seconds, on the board's own hardware. Your machine will differ.
35
Listed on the model card. The accuracy above is English only.
Percentage of words wrong on each set, lower better. Bars are scaled to this model's own worst case; the placing beneath each rate is against every model measured on that set.
Which languages ↓ ↑
English · Spanish · German · French · Italian · Arabic · Japanese · Korean · Portuguese · Russian · Hindi · Chinese · Vietnamese · Hebrew · Dutch · Czech · Danish · Polish · Norwegian · Swedish · Thai · Turkish · Bulgarian · Greek · Estonian · Finnish · Croatian · Hungarian · Lithuanian · Latvian · Romanian · Slovak · Ukrainian · Maltese · Slovenian
The figures above come from the Open ASR Leaderboard, an independent public test that runs every model on the same recordings. It is the only measurement of transcription quality we know of, so there are no other scores to show.
Each of these is the same transcription job on a different kind of recording, so together they say where it holds up and where it slips — not how closely it follows an instruction.
Every published score for this model9 scoresEvery figure we hold, from 9 boards, with who ran it and a link to the source — including the boards no rating above is built on.
Can you run it yourself?
Comfortable fit
GeForce RTX 4090 · 24 GB
Room to spare. 21.2 GB spare means a 10% error in the size would not change the answer.
Apple M1 Pro (16-core GPU) · 32 GB
Room to spare. 22.4 GB spare means a 10% error in the size would not change the answer.
Apple M1 (8-core GPU, 8GB unified) · 8 GB
Room to spare. 4.4 GB spare means a 10% error in the size would not change the answer.
These cards answer whether Nemotron 3.5 ASR Streaming 0.6b loads, not how fast it transcribes. Throughput figures for a transcription model come from its text decoder, so treat this as a fit answer rather than a speed one.
Memory use by level
Against a 24 GB card.
What is quantisation? →This model on every device we track71 devicesThe Q4 build most people download, on each device: what the weights come to, how much context the memory leaves, and whether it runs. Smallest device that runs it first. This is a fit answer: whether it loads, not how fast it transcribes.
When we formed this view
Recent changes
What moved
first indexed by our pipelineEach date is the day we first saw the change, or the day the maker announced it.
What we do not know about this model yet
- We hold no measured file for it, so all 3 sizes on this page are calculated from the parameter count.
- Nothing we hold says whether an endpoint streams, so we do not show it either way.
Licence and identifiers
What the licence allowsCustom licence, what it allows commercially, and the identifiers you need to pull this model — its Hugging Face repo, our slug and a machine-readable card.
Licence
Custom licence
This model ships custom license terms that don't map to a known template. We haven't parsed them, so commercial use, redistribution and derivatives are unverified — review the original terms before shipping.
Identifiers
- Hugging Face
- nvidia/nemotron-3.5-asr-streaming-0.6b
- Architecture
- Dense
- Takes in, gives back
- Audio in, text out
- Catalogue slug
- nvidia-nemotron-3-5-asr-streaming-0-6b