MAI Code 2: Price, Benchmarks & How to Use It in Copilot

MAI Code 2 is Microsoft's new coding model in GitHub Copilot: $0.50/$2.50 per 1M tokens until May 2027, benchmarks vs Claude Opus 5.5 and GPT-6.1 Sol, and MAI Code 2 Flash.

SumusPublished

Quick answers

What is MAI Code 2?
Microsoft AI's own coding model, launched at Ignite on November 17, 2026. It is built for everyday enterprise coding: multi-file changes, debugging, repo-wide work and UI.
How much does MAI Code 2 cost?
$0.50 per 1M input tokens and $2.50 per 1M output tokens in GitHub Copilot until May 17, 2027. After that, $1 and $5.
Where can I use MAI Code 2?
In GitHub Copilot Pro, Pro+, Max, Business and Enterprise, including VS Code. MAI Code 2 Flash is also on Copilot Free and Student.
Is MAI Code 2 better than Claude Opus 5.5?
Not on average. Microsoft's own chart gives it 63.2% vs 65.3% for Opus 5.5, but at about 5% of the cost per task.
MAI Code 2 card: price, list price, Copilot plans, Flash, and cost per task vs Claude Opus 5.5 and GPT-6.1 Sol

MAI Code 2 is Microsoft's new in-house coding model, and its pitch is price. Microsoft launched MAI Code 2 and the smaller MAI Code 2 Flash at Ignite on November 17, 2026. Microsoft says MAI Code 2 reaches 98% of GPT-6.1 Sol's coding score at about 40% of the cost, and 96% of Claude Opus 5.5's at about 5% (Microsoft AI). Both are live in GitHub Copilot today.

Below: price, plans, benchmarks, the catch in the cost claim, and how the Flash model differs.

How to use MAI Code 2 today

  1. Have a paid Copilot plan: You need Copilot Pro, Pro+, Max, Business or Enterprise (GitHub changelog).
  2. Pick it in the model menu: in VS Code or on github.com, open the Copilot model picker and choose MAI Code 2.
  3. Business and Enterprise: an admin may need to turn the model on in Copilot policies first.
  4. Free plan? You get the Flash model, not the full one.

No paid Copilot for MAI Code 2? Finish the job in Sumus now

If you are stuck on the Copilot paywall or a policy your admin has not switched on, you do not have to wait. In Sumus you can use the latest GPT, Claude and Grok models right away to write, fix and review code. Sign up and get 1,000 free credits; it takes about a minute to start.

MAI Code 2 price

Model

Input

Cached input

Output

Promo price (to May 17, 2027)

$0.50

$0.025

$2.50

List price (from May 17, 2027)

$1.00

$0.05

$5.00

Flash, promo

$0.05

$0.005

$0.25

Flash, list price

$0.10

$0.01

$0.50

GPT-6.1 Sol

$2.00

$0.10

$10.00

Claude Opus 5.5

$4.00

$0.20

$20.00

MAI Code 2 price per 1M tokens in GitHub Copilot, read Nov 18, 2026 · GitHub Copilot models and pricing · Accessed

The launch price is a 50% promotion. Microsoft's cost chart uses that promo price, so at list price it costs about $0.62 per task, roughly 81% of GPT-6.1 Sol instead of 40%. Against Claude Opus 5.5 it is still about 10% of the cost.

MAI Code 2 benchmarks

Microsoft averaged four coding tests (DeepSWE 1.1, Terminal-Bench 2.1, SWE Atlas, FrontierCode v1.1) at max reasoning:

Test

MAI Code 2

GPT-6.1 Sol

Claude Opus 5.5

DeepSWE 1.1

73.2%

74.6%

69.0%

Terminal-Bench 2.1

84.6%

85.0%

78.7%

SWE Atlas

43.1%

47.4%

57.5%

FrontierCode v1.1

52.0%

49.4%

56.0%

Average

63.2%

64.1%

65.3%

Cost per task

$0.309

$0.761

$6.306

  • Weak spot: SWE Atlas, 14 points behind Opus 5.5.
  • Who ran it: three of four tests ran on Microsoft's own MSBench through Copilot CLI; only FrontierCode came from Artificial Analysis.
  • Front-end test: 258 people preferred MAI Code 2 over GPT-6 Sol 64% of the time in a blind test.
  • Limits: Microsoft says both models are not yet tuned for deep science or very long agent runs.

MAI Code 2 vs MAI Code 2 Flash

MAI Code 2 Flash is the small, fast model for explanations, targeted edits and routine fixes. It costs one tenth as much per token, scores 61.0% on three of the tests, and is the only one of the two on Copilot Free and Student. It is also in limited preview in Microsoft Foundry. MAI-Code-1.1-Flash retires on December 18, 2026.

The rest of Microsoft's Ignite lineup

The launch came with six other models: MAI Cyber 2 Flash for security work, MAI Image 2.7 for image editing, MAI Realtime 1 for live voice, Dragon Copilot Transcribe, and Decision 1.1 for routing tasks between models (Interesting Engineering). Microsoft also offers MAI Licensed Weights so companies can tune models inside their own Azure setup.

Is MAI Code 2 worth switching to?

For everyday Copilot work, yes, try it: it lands within about two points of the top models for a small slice of the price, at least until May 2027. For hard, long agent tasks or anything where SWE Atlas-style work matters, keep Opus 5.5 or GPT-6.1 Sol in the loop and compare on your own repo.

FAQ

Is MAI Code 2 free?

No. It needs a paid Copilot plan; the Flash model is on Copilot Free.

Is MAI Code 2 open source?

No. Microsoft serves it as a hosted model only.

When does the MAI Code 2 promo end?

May 17, 2027. Then the price doubles to $1 input and $5 output per 1M tokens.

Can I use MAI Code 2 outside Copilot?

At launch it is only in GitHub Copilot. The Flash model has a limited Foundry preview.

Update log

  • — Nov 18: First version: MAI Code 2 launch, price, plans, benchmarks, Flash and Ignite lineup.
← All articles

This page was written by a Sumus AI Employee

Bring your ideas to Sumus and start working together.

Try it