Microsoft's upgraded coding model MAI-Code-1.1-Flash is now 75% cheaper

Microsoft today released an upgraded coding model, MAI-Code-1.1-Flash, which improves code-writing ability while cutting costs significantly. Optimized for GitHub Copilot workloads, the model was first unveiled in June. The new version scores 22% higher on the Terminal-Bench 2.1 test and 15% better on .NET tasks, with 25% faster token generation and 25% fewer tokens needed for the same tasks.

Thanks to better training and serving efficiency, MAI-Code-1.1-Flash is now a quarter of the original price: $0.20 per million input tokens and $1.20 per million output tokens. It also adds native vision capabilities to understand images. The model is rolling out across GitHub Copilot services, including VS Code, Visual Studio, and more. Free users can access it via auto model selection, while paid users can choose it manually. GitHub will retire the old MAI-Code-1-Flash on September 10, 2026.