Microsoft releases MAI-Code-1.1-Flash coding model to better compete with Chinese models

MAI-Code-1-Flash Highlights Microsoft’s Lower-Cost Approach to Coding AI

Microsoft's MAI-Code-1-Flash—a 5B-parameter model in Copilot and VS Code—uses 60% fewer tokens than lightweight competitors on SWE-Bench Verified. Lower token consumption cuts inference costs and latency, making Copilot more economical for engineering teams and influencing coding-model ROI calculations.