Mistral Launches Voxtral 3 Voice-Agent API Starting at $0.04 Per Minute

Mistral AI launched Voxtral 3 on Oct. 20, pairing new speech models with a voice-agent API that lets developers build systems able to listen, respond and use connected tools during calls.

The agent service costs $0.04 per minute when Mistral Small 4 handles reasoning. The rate rises to $0.08 per minute with another reasoning model. Both prices cover the conversational workflow rather than a single speech function.

The service converts incoming speech to text, passes the request to a reasoning model and returns a spoken answer. During the conversation, the model can use connected tools to retrieve information or carry out a task. Mistral’s software development kit supports phone connections through SIP and browser or app connections through WebRTC.

Developers can also use the speech models separately. Batch transcription with Voxtral Mini Transcribe 3 costs $0.0035 per minute, while Voxtral Mini Transcribe Realtime 3 costs $0.008 per minute. Voxtral Small Transcribe 3 costs $0.0040 per minute. Voxtral Speech 3, which generates speech, is priced at $13 per million characters.

That speech-generation rate uses a different billing unit from the full agent service. Scaled Cognition’s text-to-speech API for existing agents is priced at $9 per million characters. Mistral’s $0.04 and $0.08 per-minute rates also cover transcription and the agent workflow.

Mistral introduced its first Voxtral speech-understanding models in July 2025 and a text-to-speech model in March 2026. The new service brings speech input and output together with reasoning and tool calls, allowing an agent to interact with connected systems during a call.

The API is available through US, European and global endpoints. Organizations seeking to run the full system on their own infrastructure or in a private cloud must arrange deployment with Mistral.