MAI Code 2 in GitHub Copilot: Price, Plans and Seven Models
Microsoft put MAI Code 2 and Code 2 Flash in GitHub Copilot on November 17, 2026. Promo token rates double on May 17, 2027. Flash is on Free.

Microsoft put MAI Code 2 and Code 2 Flash in GitHub Copilot on November 17, 2026. Promo token rates double on May 17, 2027. Flash is on Free.

On November 17, 2026, Microsoft AI published Seven new world class models at scalable price-performance ↗ and MAI Code 2 and Code 2 Flash ↗. The GitHub changelog posted MAI Code 2 and MAI Code 2 Flash in GitHub Copilot ↗ the same day. Thurrott.com ↗ (Paul Thurrott, November 17) covered the keynote. The picture above is an original graphic created by aitoolranked. We have not used these models.
Microsoft says the coding pair is "explicitly not yet tuned for harder problems, like fundamental science research and very long horizon agentic work." GitHub's changelog calls that same pair "complex, agentic software engineering" and "designed for agentic coding workflows."
Thurrott quotes Microsoft AI CEO Mustafa Suleyman: "Our commitment is to give you the very best price performance out there so you can really scale your agents … We don't just hill-climb on frontier quality alone. We co-optimize quality and price together."
Microsoft's Ignite lines: MAI Code 2 reaches 98% of GPT-6.1 Sol's coding quality at about 40% of the price, and 96% of Claude Opus 5.5 at 5%. MAI Code 2 Flash matches GPT-6 Luna at 65% of the price and Claude Haiku 5.5 at 20%. The footnote averages DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas, and FrontierCode v1.1 at max reasoning effort. Three of four ran on MSBench, Microsoft's platform, with Copilot CLI and no subagents. FrontierCode comes from Artificial Analysis. Code 2's cost, and Flash's, is the 50% promo price. Flash's average drops FrontierCode. Reviewers preferred MAI Code 2 over GPT-6 Sol 64% of the time in Microsoft's blind frontend test, 258 annotators.
GitHub's pricing docs ↗ list these rates per million tokens. Both rows are promotional until May 17, 2027, then each rate doubles.
| Model | Input | Cached input | Output | From May 17, 2027 |
|---|---|---|---|---|
| MAI Code 2 | $0.50 | $0.025 | $2.50 | $1.00 / $0.05 / $5.00 |
| MAI Code 2 Flash | $0.05 | $0.005 | $0.25 | $0.10 / $0.01 / $0.50 |
The changelog puts MAI Code 2 on Copilot Pro, Pro+, Max, Business, and Enterprise, and Flash on those plus Copilot Free and Student, under usage-based billing. It also says: "Both models are available to eligible GitHub Copilot subscribers on annual plans." Select them in VS Code, Visual Studio, Copilot CLI, the Copilot cloud agent, the GitHub Copilot app, github.com, GitHub Mobile on iOS and Android, JetBrains, Xcode, or Eclipse. Business and Enterprise admins use the model policy. New models turn on unless an admin disabled the global default or that model.
Code 2 Flash is a limited preview in Microsoft Foundry. Microsoft says a release on Open Router, Vercel, Hermes, Open Code, and Kilo is coming soon, with no date, and that Code 2 Flash was optimized further into MAI Work Flash for Microsoft 365 and Fabric IQ.
On November 18 at 09:34 UTC (14:34 PKT), GitHub Status posted ↗: "The issues with our upstream model provider have been resolved, and MAI Code 2 Flash is once again available in Copilot products and IDE surfaces."
Still Microsoft's claims. Cyber 2 Flash beats GPT-6 Astra's quality on CyberGymE2E at under 2% of the cost per task, and output tokens are 86% below Cyber 1.1 Flash. Separate product: GPT-6 Cyber. Image 2.7 is #1 on Artificial Analysis image editing, ahead of GPT Image 2.6 Sunburst (max) at one fifth of the price, and #4 on text-to-image. Image 2.7 Flash is about $10 per 1,000 images, both in Foundry. Realtime 1, the first live speech-to-speech model, outperformed GPT-Realtime 2.1 in Microsoft's tests from $0.01 per audio minute. Try it in MAI Labs. Dragon Copilot Transcribe, on MAI Transcribe 2, covers 10 million encounters a month and halves multilingual error rates. Thurrott notes it has no post of its own.
Microsoft-Decision-1.1 ↗ is generally available in Foundry, on OpenRouter, and on Vercel. Ignite says 50x faster and 79x cheaper than GPT-Sol 6.1. The GA post says more than 4,000 customers, an open-weight Gemma 4 26B A4B base from Google that Microsoft then trained, and image input. Foundry Global Standard is $0.042 per million input tokens, output free. Gemini Decision 1 is Google's router. Microsoft also says Maia 200 is in production at 1.7x more tokens per watt, Licensed Weights run a customized model in the customer's Azure, and an AT&T model beat GPT-6.1 Sol at 50x lower cost.
"The bigger story isn't another model launch. It's Microsoft controlling both the model and its distribution." — Kiran ↗ (@KiranAIGrowth)
"MAI Code 2's "40% of the cost" is a promo price that doubles next May, and 3 of its 4 benchmarks ran on Microsoft's own eval platform. Happy users, harder comparisons." — @fierce788 ↗
"microsoft invested billions into openai, partnered with anthropic, and still just launched its own in-house coding model (mai code 2) today." — Pradyum M. ↗ (@pradyum1916)
"Benchmark it now, budget for the later rate." — AI Flood ↗ (@aiflood)
Reddit was not reachable for this story.
Try Flash for routine edits, Free and Student included. The table doubles on May 17, 2027.
Read 98% and 96% as Microsoft's. Three of four Code 2 benchmarks ran on MSBench at the promo price, and Microsoft says the models are not tuned for very long-horizon agentic work. Keep a frontier model for hard tasks. Rates: the GPT-6.2 note. Editors: Copilot vs Cursor. Field: the coding guide. The student guide predates Flash on Free. Shelf: AI Coding.
Open the picker on a monthly seat. The changelog names the plans, then eligible annual subscribers, and never describes monthly billing. Open Router, Vercel, Hermes, Open Code, and Kilo are still "soon," with no date.
Verdict: Flash for everyday edits, with May 17, 2027 beside the rate. Code 2 competes with Microsoft's partners in that picker.