- Make MiniMax-M3 the default MiniMax model - Verify the existing OpenAI-compatible adapter sends M3 requests correctly - Update AI documentation and release notes for #431