Questions? hello@viberation.devGet supportBlogDocsChangelog
Get started

Voxtral

Mistral's open speech model — audio in, text out.

Voxtral is Mistral's open-weight audio model: it takes speech and files as input and returns text, handling transcription, translation and questions about what was said, while keeping the text ability of the Mistral Small model it is built on. It is a single specialised model rather than a family, so reach for it when your app has recordings to understand, and use a general model for everything else.

At a glance — Voxtral Small 24B 2507

Cost
$0.10 in · $0.30 out per 1M tokensBudget
Memory
33K tokens — about 49 pages
Reads
Text · Files · Audio
Can do
  • Use tools (agents)
  • Structured output
  • Thinking mode (not supported)
Reliability
99.49% uptime, last 24h
Released
Oct 30, 2025
Advanced details
API model ID
mistralai/voxtral-small-24b-2507
Max output
26K tokens
Cached input
$0.01 per 1M
Tokenizer
Mistral

Providers

The same model, hosted by different companies. OpenRouter picks one per request and falls back to the next if it fails.

ProviderInput /1MOutput /1MCached /1MMax outputUptime 24h
Mistral · zdr$0.10$0.30$0.0126K99.42%
Mistral$0.10$0.30$0.0126K99.49%
Mistral · eu$0.11$0.33$0.01126K—

Supported parameters

  • frequency_penalty
  • max_tokens
  • presence_penalty
  • response_format
  • seed
  • stop
  • structured_outputs
  • temperature
  • tool_choice
  • tools
  • top_p

Live from OpenRouter, refreshed hourly.

Key info

Pricing
Open source
Category
Models
Best for
Expert
Made by
Mistral
Weights
Open
Good for
Transcribing and understanding speech

More in Models

  • France's frontier lab, with the small models everyone self-hosts.

    Models
  • Meta's current line, built for long multi-agent runs.

    Models
  • The open-weight family everyone else gets benchmarked against.

    Models
  • Google's open models — Gemini's research, small enough to self-host.

    Models