Voxtral Mini 3B 2507
Mistral AI · released Jul 1, 2025 · mistralai/Voxtral-Mini-3B-2507
Speech to textTranscribes a recording into words
- Type
- Open weightsApache License 2.0
- Languages
- 8
- Size
- 4.7B
Context measured in tokens
Our take
Written Sep 4, 2026Voxtral Mini is a compact downloadable speech-to-text model from Mistral AI that turns audio into written words. It runs very fast and carries a permissive Apache licence, but its accuracy is uneven across audio types and no commercial hosts currently offer it.
Pick this for fast local transcription of accented earnings calls, where it scores better than most alternatives, or for multilingual pipelines across its eight supported languages. Use it when you need truly open weights you can fine-tune or redistribute. Skip it if your audio is clean read-aloud, podcasts, or meeting-room recordings, where it trails the field, or if you want a managed API rather than self-hosting.
The case for it
- Better than most on accented earnings-call audio, with a lower error rate than the field median.
- Very fast: roughly an hour of audio in 20 seconds on the benchmark harness.
- Apache 2.0 licence allows commercial use, fine-tuning and redistribution.
- Eight languages supported, though only English has measured accuracy data.
The case against it
- Worse than most on clean read-aloud, podcasts and video, and meeting-room audio.
- No commercial hosting options currently listed; you must self-host or run locally.
- Release date undisclosed, and no active-parameter figure is provided.
How good is it?
An open transcription model for turning speech into text, though meeting recordings are where it struggles most.
- transcribing recordings of meetings in a roomRecorded meetings · 71st of 92
TranscriptionTurning speech into text3 of 5Open ASR WER · 47th of 76
94.5%
Misses roughly one word in 18, averaged over nine English test sets.
181×59th of 74
an hour of audio in 20 seconds, on the board's own hardware. Your machine will differ.
8
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 · French · German · Spanish · Italian · Portuguese · Dutch · Hindi
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. 18.3 GB spare means a 10% error in the size would not change the answer.
GeForce RTX 5090 · 32 GB
Room to spare. 26.3 GB spare means a 10% error in the size would not change the answer.
Apple M2 (8-core GPU, 8GB unified) · 8 GB
Room to spare. 1.5 GB spare means a 10% error in the size would not change the answer.
These cards answer whether Voxtral Mini 3B 2507 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 allowsApache License 2.0, 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
Apache License 2.0
Fully permissive: commercial use, redistribution, and derivatives allowed. Requires attribution and a copy of the license. Includes an express patent grant.
Identifiers
- Hugging Face
- mistralai/Voxtral-Mini-3B-2507
- Architecture
- Dense
- Takes in, gives back
- Audio in, text out
- Catalogue slug
- mistralai-voxtral-mini-3b-2507