Hill-climbing MAI models for GitHub Copilot and Excel
Microsoft shares results of specialized MAI models in Copilot and Excel.
Model update
Microsoft announced that its MAI-Code-1-Flash model, deployed in 
GitHub Copilot since June, has an approximately 10% higher code accept rate than GPT 5.4 Mini and Claude Haiku 4.5 in VS Code, with 10% lower median token usage.
- A specialized MAI model trained from that checkpoint for Excel is on par with GPT-5.6 for common tasks while being more cost-efficient and deployable on older GPU classes.
- Microsoft is extending this hill-climbing approach to other agentic products including Copilot Chat, Outlook, and PowerPoint.
