Vibitely Daily

Wednesday, 18 November 2026, story 2 of 7

Microsoft's MAI Code 2 models arrive in GitHub Copilot

Microsoft AI says MAI Code 2 reaches 96% of Claude Opus 5.5's performance at 5% of the cost, with cost shown at a promotional price.

Illustration: A balance scale holding a large pale block on one pan and a small dark block on the other, level, with a half-peeled price tag shape hanging from the beam.

What happened

GitHub has added MAI Code 2 and MAI Code 2 Flash, two Microsoft AI coding models, to the Copilot model picker. Code 2 is open to Pro, Pro+, Max, Business and Enterprise; Flash also covers Free and Student. Both are billed under usage-based billing, and GitHub says they are for eligible subscribers on annual plans.

Microsoft AI says Code 2 reaches 96% of Claude Opus 5.5's performance at 5% of the cost, averaged over four benchmarks. These are the company's figures, with cost shown at a 50% promotional price.

What it means

The price claims are the part to test. Microsoft AI's cost ratios use a 50% promotional price, and three of the four benchmarks ran on its internal platform with Copilot CLI and without subagents. Teams on usage-based billing should compare cost per accepted change on their own repositories before moving work over.

Microsoft AI also says the models are not yet tuned for fundamental science research or very long horizon agentic work. The pitch is everyday tasks, with Flash aimed at high-volume edits. Admins on Business and Enterprise should note that new models are enabled automatically under default enablement unless switched off. Watch for the promised releases on Open Router, Vercel, Hermes, Open Code and Kilo. The source gives no date.

What to do: Run MAI Code 2 against your current model on a few real repository tasks, and check cost at full price, not the promotional one.

Sources

Filed under Models. Published .

"What it means" is analysis, not reporting. Found an error? Write to daily@vibitely.com.