Mistral's Voxtral 3 speech models land #3 on STT leaderboard at $13/1M chars
ArtificialAnlys · x · 2026-10-21
Artificial Analysis evaluated Mistral's newly released Voxtral 3 speech family: one TTS model and three STT models.
- STT: Voxtral Small Transcribe 3 takes #3 on the speech-to-text leaderboard with a 2.1% AA-WER, ahead of ElevenLabs Scribe v2 (2.2%) and Grok Voice Transcribe 2.0 (2.3%); Mini Transcribe 3 ranks #5 at 2.2%.
- Streaming STT: Mini Transcribe Realtime 3 scores 3.3% WER at the 960ms setting (#6 of 46) and 3.6% at 480ms, both far better than the previous generation's 5.2%.
- TTS: Voxtral 3 TTS ranks #12 in both TTS arenas (Provider Voice Elo 1218, Controlled Voice 1068); pronunciation robustness is 84.4% (#13), with Standalone Terms at 93.9% (#3).
- Speed and price: 112.3 chars/sec generation, well above Eleven v4 Turbo (78.1) and Gemini 3.8 Flash TTS (55.3); priced at $13/1M characters, cheaper than Gemini ($16.5), Breeze TTS 2 ($34) and Eleven v4 Turbo ($40).
More from Models
- 3 months testing whether Anthropic's internal research model can design scorpion antivenom — apoorvasriniva · 2026-10-21
- Space Bunny Alpha revealed as MiniMax M3.1 Flash: 1M context, budget-friendly coding — ZabihullahAtal · 2026-10-21
- 'Opus 5.5 Medium' Spotted in Rork's Model Picker, Hinting at New Anthropic Model — rudrank · 2026-10-21
- RSI-Bench Clears First Task Set Through Reward-Hacking Detection Pipeline — SRSchmidgall · 2026-10-21
- Dev complains Anthropic's bio safeguards trigger just from reading a bio paper — robleclerc · 2026-10-21
- Redditor claims uncensored local DeepSeek build answers virus-writing prompt with zero hesitation — Only_Light_9817 · 2026-10-21