MAI Code 2 has exactly one official route: GitHub Copilot. This article answers three questions: whether your plan covers it, where to find it, and how to choose between MAI Code 2 and Flash — plus the cases Microsoft itself says it is not for. No standalone API, no weights, no other route.
| Copilot plan | MAI Code 2 | MAI Code 2 Flash |
|---|---|---|
| Free | Not available | Available |
| Student | Not available | Available |
| Pro | Available | Available |
| Pro+ | Available | Available |
| Max | Available | Available |
| Business | Available (admins must allow it in the model policy; enabled by default) | Available |
| Enterprise | Available (admins must allow it in the model policy; enabled by default) | Available |
All of the above comes from GitHub's Nov 17 changelog. Two further limits: both models are limited to eligible Copilot subscribers on annual plans (the changelog's wording, which does not explain what "eligible" means); and under Business and Enterprise plans, admins can switch MAI models off in the Copilot model policy — if MAI Code 2 is missing from your model picker, ask your admin whether it is allowed before blaming your plan.
GitHub's changelog confirms both models in the model picker at all of these entry points: VS Code, Visual Studio, Copilot CLI, cloud agent, GitHub.com, mobile, JetBrains, Xcode, Eclipse. They fall into three kinds of use:
In the editor (VS Code / Visual Studio / JetBrains / Xcode / Eclipse): the main route. Pick MAI Code 2 or Flash in the Copilot Chat model picker, for the full everyday coding flow: production code, deep debugging, multi-file changes, repo-wide reasoning, UI work. Flash suits code explanations, targeted edits, routine fixes and rapid iteration.
In the terminal (Copilot CLI): call it directly from the terminal, as a command-line assistant; Microsoft's four benchmarks were themselves run via the Copilot CLI (no subagents), so this is a use the model has been tested in.
Cloud agent (GitHub.com, mobile): hand tasks to the cloud agent, suited to agent-style work such as multi-file changes and repo-wide reasoning — but mind Microsoft's boundary: it is not tuned for very-long-horizon agentic work, so break long tasks into stages and check each one rather than firing off one huge task and walking away.
Copilot Fusion Standard: a second GitHub changelog entry on Nov 17 confirms MAI-Code-2 has joined Copilot Fusion Standard. Fusion routes tasks across several models, so picking Standard may land you on MAI Code 2; to pin it down, pick the model explicitly in the model picker.
| Task type | Pick | Why |
|---|---|---|
| Production code, deep debugging, multi-file changes, repo-wide reasoning, UI work | MAI Code 2 | The general-purpose flagship; Microsoft reports 98% of GPT-6.1 Sol (Microsoft-chosen benchmarks, 50% promo-price basis) |
| Code explanations, targeted edits, routine fixes, rapid iteration | MAI Code 2 Flash | The lightweight tier; Microsoft reports it comparable to GPT-6 Luna and Claude Haiku 5.5 at 65% and 20% of the cost (50% promo-price basis) |
| High-volume work (dozens or hundreds of calls a day) | MAI Code 2 Flash | Average cost per task $0.052 (Microsoft blog chart, 50% promo-price basis), about a sixth of MAI Code 2's |
| Free / Student plans | Flash only | MAI Code 2 does not cover Free and Student |
| Dashboards and visualizations | MAI Code 2 | Microsoft says MAI Code 2 is tuned for dashboards and visualizations; the 258-person front-end blind test preferred it over GPT-6 Sol (but the figure has three versions — 64% / 67% / 68% — unreconciled) |
One simple rule: the more complex the task and the wider the change, the more you lean toward MAI Code 2; the smaller, more repetitive and higher-volume the task, the more you lean toward Flash. Both share the same Copilot entry points, so switching costs one click in the model picker — split by task type first, then adjust on real results.
First, fundamental science research. Microsoft says plainly that both models are "explicitly not yet tuned" for fundamental science research. Don't count on them for it; use a stronger general model instead.
Second, very-long-horizon agentic work. The same Microsoft wording: not tuned for very-long-horizon agentic work. Break long tasks into stages and check each stage; don't expect it to run a huge agent flow end to end.
Third, deployment outside Copilot. No standalone API, no weights, and no MAI Code 2 in Foundry (Flash only, limited preview), so there is currently no official route to call it outside Copilot. Microsoft says OpenRouter, Vercel and other channels are "coming soon" (wording incomplete, unconfirmed whether live); go by what each channel actually shows.
1. Check your plan: Pro and above get MAI Code 2; Free / Student get Flash only; Business / Enterprise users should confirm their admin has allowed MAI models. 2. In VS Code (or your usual editor), open Copilot Chat and pick MAI Code 2 (complex tasks) or Flash (everyday small tasks) in the model picker. 3. Try it first on a multi-file change you know well, and compare against the model you used before — focusing on debugging depth and change completeness, the home ground Microsoft claims. 4. Watch the bill: both models bill at the 50% promotional price, so the bill changes when the promotion ends; enterprise users should budget against the post-promo relative costs (see our pricing guide). 5. Mind the retirement date: MAI-Code-1.1-Flash retires on Dec 18, 2026 — move any workflow still using it to the latest MAI Code Flash model ahead of time.
Reference: GitHub Changelog — MAI Code 2 and MAI Code 2 Flash in GitHub Copilot · GitHub Changelog — Copilot Fusion updates (MAI-Code-2 joins Fusion Standard) · Microsoft AI — Product post "MAI Code 2 and Code 2 Flash: Everyday coding at scalable prices" (rewritten and organized by ModelBrief; facts per official sources)