Microsoft launched seven of its own AI models at Ignite on 17 November 2026, and the two that matter most to working developers are already sitting in your editor. MAI Code 2 and MAI Code 2 Flash are generally available in GitHub Copilot right now, and Microsoft claims near-frontier coding quality at a fraction of the price. I spent the morning reading the announcement and the fine print, so here is the honest version: how to turn them on, what they cost, and where Microsoft own numbers need a second look.
What Microsoft claims
- MAI Code 2 reaches 98% of GPT-6.1 Sol coding quality at about 40% of the price, and 96% of Claude Opus 5.5 performance at just 5% of the price, averaged across four benchmarks: DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1.
- MAI Code 2 Flash matches GPT-6 Luna performance at 65% of the price, and Haiku 5.5 at 20% of the price.
- Where they live: both models are in the GitHub Copilot model picker and through Copilot Fusion. Flash is also in a limited-access preview in Microsoft Foundry.
- Who gets what: Code 2 is on Copilot Pro, Pro+, Max, Business and Enterprise. Flash is on all of those, plus Copilot Free and Student.
Before you get excited, two caveats from Microsoft itself. First, every figure comes from Microsoft own testing, with no independent verification yet. Second, the cost comparisons use a 50% promotional price, so the gap will narrow when the promotion ends. Microsoft also says plainly that these models are not yet tuned for the hardest problems, such as fundamental science research or very long-horizon agent work. For everyday enterprise coding, though, the claim is striking: 96% of the best Claude model for 5% of the money.
How to switch it on in 5 minutes
Step 1: check your plan. Open github.com, go to Settings, then Copilot, and confirm your plan. Free and Student users will only see Flash. Pro, Pro+, Max, Business and Enterprise users get both models.
Step 2: pick the model in VS Code. Open Copilot Chat, click the model picker at the top of the chat panel, and select MAI Code 2 for deep work or MAI Code 2 Flash for quick edits. The same picker appears in Visual Studio, on GitHub.com, in Copilot CLI, in the cloud agent and in the mobile app.
Step 3: set your default per task. Use Code 2 when the job spans multiple files, needs a refactor, or requires reasoning across a whole repository. Use Flash for explanations, targeted edits, routine fixes and rapid iteration. Switching takes one click, so there is no reason to use one model for everything.
Step 4: watch your billing. Both models use usage-based billing on eligible plans. On Business and Enterprise plans, new models are switched on by default unless an admin has turned them off, so team leads should check the admin setting before a whole team starts burning usage.
Step 5: test it on your own code, not a demo. Benchmarks are Microsoft own. The only test that matters is yours: take a real bug or a real refactor from your current project, run it through Code 2, then through your usual model, and compare the diff, the number of revision rounds, and the time it took. If Code 2 needs fewer rounds, it is already saving you money even before the price difference.
When Flash is the smarter pick
If you are on Copilot Free or Student, Flash is not the consolation prize, it is the whole offer, and it is a good one: GPT-6 Luna level performance at 65% of the price, free to try. Even on a paid plan, I would use Flash for the high-volume, low-stakes work that fills most of a coding day: explaining a function, writing a test, fixing a typo-level bug. Save Code 2 for the jobs where a wrong answer costs an afternoon. That split is exactly how Microsoft designed the pair, and it is how you keep a usage-based bill under control.
One more honest note for freelancers: a cheaper model only saves money if it finishes the job in fewer attempts. Track your first week with Code 2 the way you would track any new tool: tasks completed, revision rounds, and the usage line on your bill. If the numbers hold up, this is the cheapest near-frontier coding you have ever had. If they do not, the switch back is one click away.
Sources: Microsoft AI, Seven new world class models at scalable price-performance, 17 November 2026, microsoft.ai. Related reading on this blog: Anthropic Releases Claude Fable 5.2 and Mythos 5.2: Cheaper, Faster, and an RSI Debate and Meta Muse Ember: The $0.10 Model Built to Run Your AI Agents Grunt Work.
