🇺🇸 English 🇰🇷 한국어 🇯🇵 日本語
Home › LLM & Chatbots
LLM & Chatbots

Microsoft's MAI Code 2 Undercuts GPT-6.1 Sol

Microsoft AI released seven in-house models at Ignite on Nov. 17, 2026, led by MAI Code 2, which it says matches 98% of GPT-6.1 Sol's coding performance at about 40% of the cost. The models are available in GitHub Copilot; the results are Microsoft's own.

Microsoft AI on Tuesday, Nov. 17, 2026, used the opening day of its Ignite conference in San Francisco to release MAI Code 2, a coding model it says delivers 98% of OpenAI GPT-6.1 Sol's performance at about 40% of the cost. The model headlines a batch of seven new in-house MAI models, released alongside a specialist security model and a plan to let companies run MAI weights inside their own Azure environments.

Microsoft sign standing outside a glass office building on the company's campus
A Microsoft sign outside Building 99 on the company's Redmond campus. File photo · 사진: Robert Scoble / Flickr(CC BY 2.0)

Key takeaways

  • Microsoft says MAI Code 2 reaches 98% of GPT-6.1 Sol's score at 40% of the cost, and 96% of Claude Opus 5.5's score at 5% of the cost, averaged over four coding benchmarks.
  • MAI Code 2 and the lighter MAI Code 2 Flash are available now in GitHub Copilot; Code 2 Flash is in limited-access preview in Microsoft Foundry.
  • MAI Cyber 2 Flash scored 92.3% on CyberGym-E2E against 91.5% for GPT-6 Astra at under 2% of the cost per task, but it will ship only inside Microsoft security products.

MAI Code 2 benchmarks and pricing claims

According to Microsoft AI's launch post, the comparison averages four benchmarks that mirror everyday enterprise work seen in GitHub Copilot: DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1. All models ran at maximum reasoning effort.

The footnotes matter. Microsoft priced MAI Code 2 at its 50% promotional rate when computing cost, and it ran three of the tests on its own internal evaluation platform. The company also says both models are "explicitly not yet tuned" for harder problems such as fundamental science research and very long-horizon agentic work. No independent lab has yet replicated the results.

MAI Code 2 Flash targets high-volume tasks such as code explanations and routine fixes. Microsoft says it performs comparably to GPT-6 Luna at 65% of the cost and to Claude Haiku 5.5 at 20%. The company says the models will also come to OpenRouter, Vercel, Hermes, OpenCode and Kilo soon.

MAI Cyber 2 Flash and security pricing

Microsoft's security model lists at $0.10 per million input tokens and $0.50 per million output tokens, compared with $10 and $50 for GPT-6 Astra on Azure OpenAI. It has a one-million-token context window. Microsoft said 92 of CyberGym-E2E's 905 tasks overlapped with its training data, so it reported results for every model on the remaining 813.

The model is exclusive to Microsoft security products. In Codename MDASH, Microsoft's agentic code-scanning system, it will replace the current model for eligible customers starting in mid-December.

The rest of the MAI lineup

The other releases, per Microsoft's overview, are MAI Image 2.7, which tops Artificial Analysis' image-editing ranking; MAI Realtime 1, a speech-to-speech model from $0.01 per audio minute; Dragon Copilot Transcribe, which handles 10 million patient encounters a month; and Microsoft-Decision-1.1, a routing model built on Google's Gemma. Microsoft added that its Maia 200 accelerator is now in production, producing 1.7 times more tokens per watt.

Why Microsoft is selling price, not peak scores

Microsoft first shipped an in-house coding model, MAI-Code-1-Flash, at its Build conference on June 2. Five months later, the pitch has shifted from catching up to undercutting. "Your IQ. Your agents. Your weights. All under your control," the MAI announcement reads. The same argument is playing out among agent makers, from Meta's Muse Code terminal agent to Google's Gemini 4 Carbon, which competes on cost per task.

What's next

Microsoft says MAI Licensed Weights, which let organizations customize an MAI model with their own data and run it in their own Azure environment, are coming soon through its Frontier Company program. Watch for third-party benchmark runs and for list pricing once the promotional rate ends.

FAQ

Where can developers use MAI Code 2? In GitHub Copilot now. Code 2 Flash is also in limited-access preview in Microsoft Foundry.

Can anyone buy MAI Cyber 2 Flash? No. Microsoft says it is coming exclusively to its own security products, including Codename MDASH.

Are the benchmark numbers independent? No. They are Microsoft's own results, and the cost figures use a 50% promotional price.

References

#Microsoft #MAI Code 2 #MAI Code 2 Flash #MAI Cyber 2 Flash #Microsoft Ignite 2026 #GitHub Copilot #Microsoft Foundry #GPT-6.1 Sol #Claude Opus 5.5 #AI coding models
S
Seung Jung
AI Newsway · Developer

Editor responsible for AI Newsway. Writes about AI, developer tools, cloud and productivity technology.

More by this reporter ›
Copyright © AI Newsway All rights reserved. Unauthorized reproduction, redistribution, and AI training prohibited.
Share
f X K

Comments 0

Comments are currently disabled.

Be the first to comment.