> ## Content Index
> Fetch the complete content index at: https://claudefox.com/llms.txt
> Use this file to discover other available public pages before exploring further.

# MAI Code 2 in GitHub Copilot: plans, per-token prices, and what the benchmark math assumes
- URL: https://claudefox.com/mai-code-2-github-copilot-pricing-plans/
- Published: 2026-11-17T21:29:54.000Z
- Updated: 2026-11-17T21:30:01.000Z
- Description: Microsoft's MAI Code 2 and MAI Code 2 Flash are now in the GitHub Copilot model picker. Here is which plans get them, what they cost per million tokens during and after the promo, and the fine print behind the '96% of Opus 5.5 at 5% of the cost' claim.
- Author: hupicake
- Tags: News

Microsoft AI released two coding models on November 17, 2026: MAI Code 2, a general-purpose model for debugging, multi-file changes and repo-wide work, and MAI Code 2 Flash, a cheaper model for explanations, targeted edits and routine fixes. Both went live in GitHub Copilot the same day. Microsoft's headline is price: it says MAI Code 2 delivers 96% of Claude Opus 5.5's performance at 5% of the cost.

That number depends on a promotional price that ends on May 17, 2027\. Below: who can use the models, what they cost per token in Copilot, and how the comparison changes once the promo ends.

## Quick facts

- Models: MAI Code 2 and MAI Code 2 Flash, from Microsoft AI (MAI).
- Where: GitHub Copilot now. MAI Code 2 Flash is also in limited-access preview in Microsoft Foundry, and Microsoft says it will 'soon' release it (the sentence follows the Flash one) on OpenRouter, Vercel, Hermes, Open Code and Kilo, with no date.
- Copilot plans: MAI Code 2 on Copilot Pro, Pro+, Max, Business and Enterprise. MAI Code 2 Flash on all of those plus Copilot Free and Student.
- Billing: usage-based, converted to GitHub AI Credits (1 credit = $0.01).
- Replaces: MAI-Code-1.1-Flash leaves every Copilot experience on December 18, 2026; GitHub names the latest MAI Code Flash model as the alternative. Enterprise admins may need to enable it in model policies.
- Not tuned for: Microsoft says the models are 'explicitly not yet tuned' for fundamental science research or very long-horizon agentic work.
- Annual plans: GitHub says both models are available to eligible subscribers on annual plans; annual Pro and Pro+ subscribers on request-based billing have different model multipliers.

## How to pick it in Copilot

Open the model picker and choose MAI Code 2 or MAI Code 2 Flash. GitHub lists the picker in VS Code, Visual Studio, Copilot CLI, the Copilot cloud agent, the GitHub Copilot app, github.com, GitHub Mobile, JetBrains, Xcode and Eclipse.

On Copilot Business and Enterprise, admins control access through the model policy in Copilot settings. Under default model enablement, new models switch on automatically unless an admin has turned off the global default or disabled the model. If you can't see it, ask whoever owns your Copilot policy.

## What it costs per million tokens

GitHub's Copilot pricing page lists these rates. Both MAI models are at promotional prices until May 17, 2027, and the page states the price from that date:

- MAI Code 2, until May 17, 2027: $0.50 input, $0.025 cached input, $2.50 output.
- MAI Code 2, from May 17, 2027: $1.00 input, $0.05 cached input, $5.00 output.
- MAI Code 2 Flash, until May 17, 2027: $0.05 input, $0.005 cached input, $0.25 output.
- MAI Code 2 Flash, from May 17, 2027: $0.10 input, $0.01 cached input, $0.50 output.
- For comparison on the same page: Claude Opus 5.5 $4 / $0.20 / $20; GPT-6.1 Sol $2 / $0.10 / $10 (up to 272K input); Claude Haiku 5.5 $0.10 / $0.01 / $0.50 (up to 100K input); GPT-6 Luna $0.10 / $0.01 / $0.50 (up to 272K input).

Per token, MAI Code 2's promo price is one-eighth of Opus 5.5 and one-quarter of GPT-6.1 Sol. At the post-promo price it is one-quarter of Opus 5.5 and half of GPT-6.1 Sol. MAI Code 2 Flash's promo price is half of Haiku 5.5 and GPT-6 Luna; after May 17, 2027 its listed rates match theirs. GitHub's Microsoft price table also has no cache-write column, while Anthropic models are listed with a cache-write cost on top of cached input.

A worked example, using the same token counts for every model (real runs differ, since models use different numbers of tokens) and default-tier rates, i.e. spread across requests that each stay under GPT-6.1 Sol's 272K and Haiku 5.5's 100K long-context thresholds: a task that sends 1M input tokens in total, 800K of them cached, and gets 50K output tokens back. Before cache-write charges, that comes to about $0.25 on MAI Code 2 at the promo price, $0.49 after it, $0.98 on GPT-6.1 Sol and $1.96 on Opus 5.5\. On MAI Code 2 Flash it is about $0.03 at the promo price and $0.05 after, the same as Haiku 5.5.

## The fine print behind '5% of the cost'

Microsoft's percentages compare the cost of its benchmark runs, which depends on how many tokens each model used, not just the price per token. Each claim has its own footnote:

- 96% of Opus 5.5 at 5% of the cost, and 98% of GPT-6.1 Sol at 40%: averaged over four benchmarks, all models at maximum reasoning effort. DeepSWE 1.1, Terminal Bench 2.1 and SWE-Atlas ran on Microsoft's internal MSBench platform with the GitHub Copilot CLI, without subagents; FrontierCode v1.1 scores and costs came from Artificial Analysis, and in that same sentence the footnote says MAI Code 2's cost is shown at its 50% promotional price.
- MAI Code 2 Flash comparable to GPT-6 Luna at 65% of the cost and Claude Haiku 5.5 at 20%: three coding benchmarks, maximum effort, Flash's cost shown at its 50% promotional price.
- The 64% preference over GPT-6 Sol on frontend work: blind side-by-side ratings by 258 annotators covering UIs, games, data visualizations, 3D scenes and SVG.

The promo price matters. For MAI Code 2 Flash, the footnote is unambiguous: at full price and the same token use, its cost would double, from 65% of GPT-6 Luna's to about 130%, and from 20% of Haiku 5.5's to about 40%. For MAI Code 2, the promo note sits in the FrontierCode sentence, so it may cover only part of the averaged figures; at full price the comparisons would rise by up to double, to as much as roughly 80% of GPT-6.1 Sol's cost and 10% of Opus 5.5's. That is our arithmetic on Microsoft's figures, not a number Microsoft published.

The 98% and 96% performance figures are averages across four benchmarks Microsoft says reflect the everyday enterprise coding it sees in Copilot. They don't cover the very long-horizon agentic work Microsoft says the models aren't tuned for yet.

## Who should try it

If most of your Copilot usage is routine edits, explanations and small fixes, MAI Code 2 Flash is the cheapest way to spend AI credits until May 2027, and it is the only one of the two on Copilot Free. If you use MAI-Code-1.1-Flash today, switch before December 18\. For long autonomous sessions, Microsoft's own positioning points you to other models.

**Try it with your AI.** Paste this into Copilot Chat (or any coding agent) to compare MAI Code 2 with your current model on your own repo before you switch:

```
I want to compare two coding models on my own repository before changing my default.

(Before pasting, confirm in the model picker that [MODEL A, e.g. MAI Code 2] is
selected.)

Then do this task without committing anything: [ONE REAL, SMALL TASK FROM MY BACKLOG,
e.g. fix issue #123].

1. Show me your plan before editing any file and wait for my OK.
2. Make the change, then run [MY TEST COMMAND] and show the output.
3. At the end, list: files changed, tests run and their result, anything you were unsure
   about.

Do not install new packages, touch credential files, or delete files. I will run the
same task again with [MODEL B, e.g. my current model] and compare the diffs, the test
results and the AI credits each run used in my Copilot billing and usage settings.
```

## FAQ

Is MAI Code 2 free in GitHub Copilot? MAI Code 2 Flash is listed for Copilot Free; MAI Code 2 needs Pro or above. Usage draws on your plan's included AI credits, and usage beyond them is billed per token.

When does the promotional price end? GitHub's pricing page says the promotional rates run until May 17, 2027, then rise to $1.00 / $0.05 / $5.00 per million tokens for MAI Code 2 and $0.10 / $0.01 / $0.50 for MAI Code 2 Flash.

Can I stop Copilot from charging me past my plan? GitHub's docs say to create a Bundled AI credits budget set to $0 and enable 'Stop usage when budget limit is reached'.

Can I use MAI Code 2 outside Copilot? Not yet broadly. MAI Code 2 Flash is in limited preview in Microsoft Foundry, and Microsoft says it will soon release it on OpenRouter, Vercel, Hermes, Open Code and Kilo; the announcement doesn't give a date or say the same for MAI Code 2.

## Related on claudefox

- [The 'Haiku builds, Opus advises' setup for Claude Code](https://claudefox.com/the-haiku-builds-opus-advises-setup-for-claude-code/)
- [Hugging Face Inference Providers now route through OpenRouter: what changes for your code, keys and Claude Code](https://claudefox.com/hugging-face-inference-providers-openrouter-routing-custom-keys/)
- [Plannotator: review your coding agent's plans, diffs and messages before it acts](https://claudefox.com/plannotator-review-your-coding-agent-s-plans-diffs-and/)
- [Microsoft shares slide after the Claude Sheets launch](https://claudefox.com/microsoft-shares-slide-after-the-claude-sheets-launch/)

Source: [Microsoft AI: MAI Code 2 and Code 2 Flash](https://microsoft.ai/news/introducing-mai-code-2-and-mai-code-2-flash/?ref=claudefox.com), [Microsoft AI: Seven new models at Ignite 2026](https://microsoft.ai/news/microsoft-ignite-2026/?ref=claudefox.com), [Microsoft AI: MAI Code 2 model page](https://microsoft.ai/models/mai-code-2/?ref=claudefox.com), [GitHub changelog: MAI Code 2 and MAI Code 2 Flash in GitHub Copilot](https://github.blog/changelog/2026-11-17-mai-code-2-and-mai-code-2-flash-in-github-copilot?ref=claudefox.com), [GitHub changelog: Upcoming deprecation of MAI-Code-1.1-Flash](https://github.blog/changelog/2026-11-17-upcoming-deprecation-of-mai-code-1-1-flash-in-december?ref=claudefox.com), [GitHub Docs: Models and pricing for GitHub Copilot](https://docs.github.com/copilot/reference/copilot-billing/models-and-pricing?ref=claudefox.com)

This story first appeared in the [claudefox daily briefing for November 18, 2026](https://claudefox.com/daily-2026-11-18/).