Microsoft put its own cheaper coding model inside GitHub Copilot
Microsoft has shipped MAI-Code-1-Flash, its first in-house coding model, live now in the GitHub Copilot and VS Code model picker, including the default auto-picker. Microsoft says it solves problems with up to 60% fewer tokens, lowering cost and latency.
What it actually means
If you're already in Copilot, there's now a cheaper option a click away, no migration. It won't out-think the top models on the hard parts, but the boring middle of most coding work doesn't need them. Try it on one real task this week and watch the token bill, not just the output; keep the expensive model for the genuinely hard problems. The wider signal: Microsoft is building its way off depending on OpenAI and pushing the price of everyday coding down as it goes.
A cheaper coding model, one click away in the picker. Try it on the boring middle.
Source: Microsoft AI · Friday, 26 June 2026