MiniMax provides an OpenAI-compatible API and their M2.7 model is worth supporting as a first-class provider.

What's needed

• from_minimax() factory in instructor/providers/minimax/
• MINIMAX_TOOLS and MINIMAX_JSON modes in Mode enum
• Wire into handle_response_model and handle_reask_kwargs in processing/response.py
• JSON mode via system prompt (MiniMax does not support response_format)
• Strip <think>...</think> tags from JSON mode output (reasoning models emit these)
• from_provider("minimax/...") support in auto_client.py
• Integration tests against real API (requires MINIMAX_API_KEY)
• Docs at docs/integrations/minimax.md

Notes

• API base URL: https://api.minimax.io/v1
• Primary models: MiniMax-M2.7, MiniMax-M2.7-highspeed
• Uses standard OpenAI SDK with custom base URL — no separate SDK needed