500xEngineer.

All postsSHORT3 min read

MAI Code 2 in GitHub Copilot: Cheaper Than Claude Opus, With a Catch

MAI Code 2's promo input and output rates are one eighth of Claude Opus 5.5 per token in Copilot. What Microsoft's tests say, and where the limits are.

Copilot users: paying Opus prices for routine fixes? Microsoft's cheaper model has a catch. MAI Code 2's promotional input and output rates in GitHub Copilot are one eighth of Claude Opus 5.5's per token. Microsoft's reported benchmark costs measure something different, and the promo has an end date.

Video not loading? Watch the Short on YouTube

What Microsoft released

Microsoft AI released MAI Code 2 and MAI Code 2 Flash on Nov 17, 2026, at Microsoft Ignite. Both are in GitHub Copilot. Manual selection in the model picker requires an eligible paid Copilot plan; Flash is also available to Copilot Free and Student users through auto model selection.

Microsoft describes MAI Code 2 as built for everyday enterprise coding, including deep debugging, multi-file implementation, repo-wide reasoning and UI. Flash is aimed at high-volume work such as code explanations, targeted edits, routine fixes and rapid iterations.

What Microsoft's tests say

Microsoft says MAI Code 2 reaches 96% of Claude Opus 5.5's performance at 5% of its cost per task in Microsoft's own evaluation. That's below Opus's score. Microsoft also reports 98% of GPT-6.1 Sol's performance at 40% of its cost per task in the same evaluation.

Those are average pass rate and cost per task across 4 coding benchmarks: DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1. All models ran at max reasoning effort.

Microsoft ran 3 of the 4 on MSBench, its internal evaluation platform, using GitHub Copilot CLI without subagents. FrontierCode scores and costs came from Artificial Analysis. MAI Code 2's cost was calculated at its 50% promotional price. Microsoft says results vary with workload, model configuration, reasoning effort and pricing assumptions.

Microsoft also says MAI Code 2 is tuned for dashboards and visualizations. In Microsoft's own blind frontend test, reviewers preferred its results over GPT-6 Sol 64% of the time. That comparison used GPT-6 Sol. The benchmark comparison above used GPT-6.1 Sol.

What it costs in GitHub Copilot

GitHub's listed rates below are per 1 million tokens. Input and output are separate:

Model Promo input Promo output Input starting May 17, 2027 Output starting May 17, 2027
MAI Code 2 $0.50 $2.50 $1 $5
MAI Code 2 Flash $0.05 $0.25 $0.10 $0.50

Claude Opus 5.5 is listed at $4 input and $20 output per 1 million tokens in the same Copilot table. MAI Code 2's promotional input and output rates are each one eighth of Opus's per token. Starting May 17, 2027, those rates become one quarter of Opus's current listed rates per token.

The reported cost-per-task comparison above depends on Microsoft's benchmark runs. Token rates alone don't tell you what a finished task costs.

Copilot uses usage-based billing in GitHub AI Credits. Plans include allowances that vary by plan, and usage beyond the allowance is billed at the listed token rates.

The catch

The headline performance numbers come from Microsoft's own evaluation, with most benchmarks run on its internal platform and costs calculated at promotional pricing. The MAI rates double on May 17, 2027.

Microsoft says these models aren't yet tuned for harder problems such as fundamental science research or very long horizon agentic work. I'd test them on routine changes in my own repo before choosing a default, and keep another model available for demanding agent runs.

Where you can use it

MAI Code 2 requires an eligible paid Copilot plan: Pro, Pro+, Max, Business or Enterprise. MAI Code 2 Flash is also available to Copilot Free and Student users through auto model selection.

On an eligible paid Copilot plan, you can select MAI Code 2 or MAI Code 2 Flash in the model picker. GitHub lists the models in the model picker for Visual Studio Code, Visual Studio, Copilot CLI, GitHub Copilot cloud agent, GitHub Copilot app, github.com, GitHub Mobile on iOS and Android, JetBrains, Xcode and Eclipse. Business and Enterprise administrators can manage access through model policy in Copilot settings.

Here's how you get paid

My opinion: sell small businesses custom dashboards for sales, bookings or inventory, built with MAI Code 2 in Copilot. Microsoft says the model is tuned for dashboards and visualizations, which makes that a specific offer I'd try.

I'd start with the client's data and the view they need, use MAI Code 2 for everyday code, and switch to Opus or another model when the hard parts stall. No income guaranteed.

Daily email

Get each breakdown in your inbox.

One email each evening, right after the Short goes live: the AI tool it covers, the full write-up with sources, and how to get paid for it. Free, and you can unsubscribe anytime.

Read past emails on Substack

Sources

All postsWatch on YouTube