whatmodel.sh / tts / Voxtral TTS (4B)

Voxtral TTS (4B)

+ contribute
Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Voxtral TTS (4B) is an open-weights text-to-speech model from Mistral, released March 2026. Its published size is 4B parameters. Licence: cc-by-nc-4.0.

71.4±3.8#1 of 118 text-to-speech modelsfrom 4 benchmark results across 1 source

Score within the Vapi Humanness · Bradley–Terry rating, 2026-09-25 snapshot comparison set only.

$ how it compares

The text-to-speech models ranked either side of Voxtral TTS (4B), out of 148 with a score.

#modelscorewhereRAMreleased
2Azure Neural TTS Xiaoxiao Microsoft Azure74±4comparison only
3Gemini 2.5 Flash Preview TTS Google72±3comparison only2025-05
4Voxtral TTS (4B) Mistral71±4api $16.00/M chars +local~10GB2026-03
5OpenAudio S1 Fish Audio71±8api $15.00/M chars2025-06
6Gemini 2.5 Pro Preview TTS Google69±3comparison only2025-05

see all text-to-speech models · find the ones that fit your machine

$ what it's good at

voice cloning█···█······█ #1 of 3 · 94% of its score · 4 tests
testresultvs the others in its rundifficultycountswhy
Seed-TTS · report split unspecified, speaker similarity lab62.8% ✎██·········█ beats 2/2hard · avg 22%▇▇▇▇▇▇▇▇▇▇▇▇▇ 33%at its level · small run
MiniMax TTS · 9-language speaker similarity lab74.3% ✎██·········█ beats 2/2hard · avg 31%▇▇▇▇▇▇▇▇▇▇▇▇ 30%at its level · small run
MiniMax TTS · 9-language WER (Voxtral ASR) lab1.9% ✎█····█·····█ beats 1/2avg 5.4%▇▇▇▇▇▇▇ 17%small run
Seed-TTS · report split unspecified (Voxtral ASR) lab1.2% ✎█··········▄ beats 1/2avg 3.3%▇▇▇▇▇▇ 15%small run

$ where to run it

Through an API

api · Mistral AI · $16.00/M chars · docs.mistral.ai ↗ · price sourceUSD per 1M input characters for Voxtral TTS March 2026 (voxtral-mini-tts-2603). Product: La Plateforme text-to-speech. API model: voxtral-mini-tts-2603. Settings: endpoint=/v1/audio/speech. Price basis: published rate. Currency: USD.

Other ways to run it

version

  1. Run it as published · needs ~10.1 GB, NVIDIA only · from huggingface.co/mistralai
    uv pip install -U vllm vllm-omni && vllm serve mistralai/Voxtral-4B-TTS-2603 --omni

Weights and documentation: huggingface.co/mistralai ↗

$ how the score was worked out · Voxtral TTS 94%

Each result implies a score on its own, after adjusting for how hard its test is and who ran it. The score is their weighted average, with a small pull toward the middle for thinly tested models. Each test's weight is its counts column above; here it's added up by who ran the tests. How it works

sourceresultsshare of the scorelicence
Voxtral TTS · lab-reported4▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇▇ 94%CC BY 4.0
prior · pull toward the middle▇▇ 6%
weighted average= 71.4

$ improve this page

Update model informationAdd a place to run itAdd a test resultPropose a benchmarkCorrect a benchmark scoreEdit data on GitHub

Use a short GitHub form or edit the data through a pull request. Every number needs a source; here's which ones we can use.

data updated · sources and licences · json