Voxtral Small

Voxtral Small is an instruct model from Mistral AI with native audio input for speech understanding and tool use. It is the first version of the voxtral line. It has a context window of 32000 tokens and a maximum output of 32000 tokens. The price starts at $0.10 per million input tokens and $0.30 per million output tokens, with cached input at $0.01 per million tokens. The weights are open. It accepts audio and text and returns text. It does not have a reasoning mode. It supports tool calling. It is served by 4 hosts.

First version of its line in the registry.

• ·Native audio input for speech understanding
• ·Tool calling for function use
• ·Open weights for self-hosting
• ·32k token context and output
• ·Accepts audio and text, returns text

• Reach for it for speech understanding from audio
• Reach for it for tool use in conversations
• Reach for it for building with open weights
• Reach for it for long context tasks up to 32k tokens

4 hosts serve this model at the price above · the maker's documentation

The registry keeps the six most recent versions of a line. Older ones stay in the file and can be surfaced without a rebuild.

What is the context window and output limit?

Both are 32000 tokens.

Does it support tool calling?

Yes, it supports tool calling.

Is it open weights?

Yes, the weights are open.