Expressive English, one_eval (internal)
We haven't written a description for this benchmark yet; see the card notes below for how each maker ran it.
- ModelScoreYour cost / mo
- 1Voxtral TTS 3 · Mistral AI3.13 /5
“Best value”: no other model here reports a higher score at a lower cost for your workload. Scores are self-reported by each maker with different settings, so treat close calls as ties.
How each maker reported it
- Voxtral TTS 3 launch post (quality-per-dollar charts, generated 20 Oct 2026)
Mistral's own evaluations with gemini-3.8-flash as judge. Voice Code Bench (BeSimple AI): 300 English customer-support prompts on numbers and names, % faithful clips. Pronunciation control: 300 sentences with a rare or made-up word and its IPA, % pronounced as intended (internal). Multilingual: 377 prompts, 11 languages, mean judge score 1–5 (internal). Expressive English (one_eval): 237 prompts, mean judge score 1–5 (internal). Higher is better. Prices in the charts: Voxtral $13, Eleven v4 $80, Eleven v4 Turbo $40, Cartesia Sonic 3.6 $50, Gemini 3.8 Flash TTS ≈$15.56 per 1M characters.
Only each model's own maker's number is shown; scores a maker reports for rival models are left out.