Mistral Function Calling And Tool Use
Mistral AI API · Mistral AI
Foundation Model & API — Mistral AI
Mistral AI evals — Function Calling & Tool Use (relift v3 InfraRed)
About Mistral AI
Mistral AI is a European foundation-model company offering open-weight and commercial models (Mistral Large, Codestral, Pixtral) via La Plateforme, plus Le Chat, embeddings, fine-tuning, and agents — with a strong emphasis on EU data residency.
Sample tests· showing 3 of 9
| # | Input | Expected behavior | Check |
|---|---|---|---|
| 01 | Workflow requires the next turn to call some tool (no plain text); the operator wants to force a tool call on /v1/chat/completions. | Set tool_choice='any' so the model must call one of the provided tools; to force a specific tool, pass tool_choice={type:'function',function:{name:...}}. Then handle the returned tool_calls[]. | Pass / FailAi Platformhigh |
| 02 | On a final summarization turn the operator wants the model to answer in prose and never call a tool, even though tools[] is still attached. | Set tool_choice='none' to force a text-only answer while keeping tool definitions in context; verify no tool_calls are returned before treating the content as the final answer. | Pass / FailAi Platformmedium |
| 03 | An assistant turn returns two tool_calls (get_weather Paris, get_weather Lyon), each with a distinct id; the agent executes both. | Return one role=tool message per tool_call_id, matched exactly by id, before the next model turn. Every tool_call must be answered; the tool message must carry the matching tool_call_id. | Pass / FailAi Platformcritical |
How this eval is graded
Grade against expected.ideal_behavior and expected.rubric. Per-criterion pass requires mean >= 4.0 and no criterion below 3.
Rubric criteria
- Mistral
- Ai Platform
- Function Calling And Tool Use
Recommended for
Works with
Related evals
Run this eval in your workspace
Connect your data, configure thresholds, and review results with your team.