Voxtral Small (24B-2507)

Mistral AI's Apache-2.0 open-weights audio-understanding models for transcription and spoken Q&A; Small is the 24B size.

Where it ranks

A
Transcription
Models · Local · 2.8% on Artificial Analysis Speech to Text: AA-WER v2 (non-streaming)

Vendor: Mistral AI
Family: Voxtral
Released: 15 Jul 2025
Weights: Open: download and run it yourself
License: Apache-2.0
Parameters: 24.26B
GPU memory to run locally: ~55 GB (Model card: running Voxtral-Small-24B-2507 on GPU requires ~55 GB of GPU RAM in bf16 or fp16.)
API price: $0.0040 per minute of audio (Artificial Analysis lists $4 per 1,000 minutes (Voxtral Small, provider Mistral); divided by 1,000.)
Access: Free (open weights)
Last verified: 18 Oct 2026

Price from artificialanalysis.ai, 18 Oct 2026.

Benchmark	Result	Entry on the leaderboard	As of	Source
Artificial Analysis Speech to Text: AA-WER v2 (non-streaming)	2.8%	`Voxtral Small` AA-WER v2 (non-streaming), provider Mistral; embedded value 0.0277 converted to percent and rounded; page shows no last-updated date.	18 Oct 2026	artificialanalysis.ai
Hugging Face Open ASR Leaderboard (English, short-form average WER)	6.4%	`mistralai/Voxtral-Small-24B-2507` Average WER over the board's default datasets (English short-form, version 17-10-2026); rank 32; read from the Space's initial table data.	17 Oct 2026	huggingface.co

No tracked tool lists this model yet; as open weights it runs in local runtimes.