
MAI
Microsoft's top-tier model family
222 followers
Microsoft's top-tier model family
222 followers
Microsoft AI is pioneering the future of what AI can do and what technology can be.
This is the 5th launch from MAI. View more

MAI Code 2 and Code 2 Flash
Launching today
MAI Code 2 and MAI Code 2 Flash are Microsoft AI’s new coding models. Code 2 handles production code, deep debugging, multi-file work and UI. Flash covers explanations, targeted edits and routine fixes at lower cost. Per Microsoft, Code 2 reaches 98% of GPT-6.1 Sol’s performance at 40% of its cost. Both are at a 50% promotional price and available in GitHub Copilot. Flash is also in limited-access preview on Microsoft Foundry.








Launch tags:Artificial Intelligence
Launch Team

Render Sandboxes: a computer for every agentGive every agent its own machine. Starts in under a second
Promoted

Microsoft AI just announced MAI Code 2 and MAI Code 2 Flash, two new coding models built for enterprise work.
MAI Code 2 is for everyday enterprise coding: taking code to production, deep debugging, multi-file implementation, repo-wide reasoning and UI work. It’s also tuned for dashboards, visualizations and visual coding. Microsoft says it produces changes that maintainers would merge.
MAI Code 2 Flash is the lighter model for high-volume work, such as code explanations, targeted edits, routine fixes and quick iterations, at a much lower cost.
Per Microsoft’s own benchmarks:
Code 2 averages 98% of GPT-6.1 Sol’s performance at 40% of its cost, and 96% of Claude Opus 5.5’s performance at 5% of its cost. That’s across DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1.
In blind frontend evaluations with 258 annotators, developers preferred Code 2’s results over GPT-6 Sol 64% of the time.
Code 2 Flash is comparable to GPT-6 Luna at 65% of the cost, and to Claude Haiku 5.5 at 20% of the cost.
Availability: both models are in GitHub Copilot, and Code 2 Flash is in limited-access preview on Microsoft Foundry. OpenRouter, Vercel, Hermes, OpenCode and Kilo are listed as coming soon. Both are at a 50% promotional price, and the page doesn’t give list prices.
Microsoft says neither model is tuned for fundamental science research or very long-horizon agentic work. The cost comparisons use the promotional price, and results vary by workload, configuration and reasoning effort, so it’s worth testing on your own codebase.
Good fit if your team already uses GitHub Copilot and wants a lower-cost model for everyday coding, with Flash for quick, high-volume edits.
Learn more: Microsoft AI announcement