Mistral AI introduces Responses API for building advanced AI agents
Mistral AI introduced the Responses API in public preview. This new API is designed specifically to simplify the process of building complex AI agents.
It fundamentally changes the approach to receiving responses: instead of a single final text, the API returns three separate, structured elements—the model's answer, its internal reasoning, and external tool calls. This allows developers to process each stage of the agent's work much more easily than when using standard Chat Completions.
Furthermore, Mistral has expanded model availability by adding DeepSeek V4.1 Flash (ideal for coding and agent tasks) and GLM 5.3 Flash. All these tools are available in the EU and the US, expanding capabilities for enterprise developers.
Why it matters
- —The Responses API significantly simplifies the development of complex AI agents by providing structured access to reasoning and tool calls.
- —Separating the response into individual elements (answer, reasoning, tool call) increases the reliability and manageability of AI systems.
- —The expansion of the model lineup (DeepSeek V4.1 Flash, GLM 5.3 Flash) gives developers more choice for different tasks.
Key facts
- The Responses API returns the answer, reasoning, and tool calls as separate elements.
- The new API is compatible with all Mistral chat models, including Large 4.
- DeepSeek V4.1 Flash is excellent for coding tasks and agent scenarios.
- GLM 5.3 Flash is now publicly available in the EU and US.
The full text is in the original source. Here we provide a brief summary and key facts.