GitHub Copilot Upgrades Its Coding Model: MAI-Code-1.1-Flash Cuts Cost 73% and Adds Vision

Microsoft's MAI-Code-1.1-Flash is now in GitHub Copilot: 73% cheaper than its predecessor, 22% better on Terminal-Bench CLI tasks, and it can now read images — MAI-Code-1-Flash retires September 10.

Microsoft released MAI-Code-1.1-Flash on August 11 as the mandatory successor to MAI-Code-1-Flash across all Copilot surfaces. The upgrade is significant on every dimension that matters for repetitive agentic workflows:

• 73% lower list price than MAI-Code-1-Flash
• 25% fewer tokens for the same output (lower effective cost on top of the rate cut)
• 25% faster streaming
• 22% better on Terminal-Bench 2.1 for Copilot CLI tasks; 15% better on .NET completions
• Native vision — the model accepts images, enabling screenshot-to-component, diagram-to-spec, and error-screenshot-to-fix directly in Copilot Chat

Paid individual, business, and enterprise customers can select MAI-Code-1.1-Flash now across Copilot CLI, Copilot Chat on GitHub.com, VS Code, Visual Studio, JetBrains, Eclipse, Xcode, and GitHub Mobile. MAI-Code-1-Flash retires on September 10, 2026.

Why it matters: The combined price and token-efficiency gains make 1.1-Flash the clear default for high-volume, repetitive Copilot CLI workflows — the kind where a coding agent calls the model hundreds of times per task. Vision support closes a gap that required pasting screenshot text into prompts; now you can pass the image directly.

Sources: MAI-Code-1.1-Flash available in GitHub Copilot — GitHub Changelog · Microsoft releases MAI-Code-1.1-Flash — Neowin