Microsoft launches MAI Code 2 and six other in-house AI models at Ignite
Microsoft AI on November 17 released seven in-house models at Ignite 2026, led by MAI Code 2, a coding model it says matches 96% of Claude Opus 5.5's results at 5% of the price. It is available in GitHub Copilot now.
Cortex, covering AI models, labs and agents
, 3 min read
5%MAI Code 2's price relative to Claude Opus 5.5, per Microsoft
What to know
- Microsoft says MAI Code 2 reaches 98% of GPT-6.1 Sol's coding results at about 40% of the price, and 96% of Claude Opus 5.5's at 5%.
- MAI Code 2 and MAI Code 2 Flash are in GitHub Copilot; Flash is also in a limited-access preview in Microsoft Foundry.
- The other releases cover security, images, live speech, medical transcription and model routing.
Microsoft AI, the company's in-house model group, released seven models at its Ignite 2026 conference on November 17. The headline is MAI Code 2, a coding model Microsoft is pitching on price against the coding models from OpenAI and Anthropic that already run inside its own products.
MAI Code 2 and MAI Code 2 Flash
MAI Code 2 is built for everyday enterprise coding: production changes, debugging, edits that span many files, and front-end work. Microsoft says it is not yet tuned for fundamental science research or very long-running agent tasks.
Across four coding benchmarks (DeepSWE 1.1, Terminal-Bench 2.1, SWE-Atlas and FrontierCode v1.1), Microsoft says MAI Code 2 reaches 98% of OpenAI's GPT-6.1 Sol at about 40% of the price, and 96% of Anthropic's Claude Opus 5.5 at 5% of the price. Securities.io and Unite.AI give the underlying figures: a 63.2% average pass rate at $0.309 per task, against 64.1% and $0.761 for GPT-6.1 Sol and 65.3% and $6.306 for Claude Opus 5.5, all at maximum reasoning effort. Both outlets say the MAI cost figure uses a 50% promotional price; Microsoft's announcement page does not mention a promotion. In a blind comparison of front-end work, Microsoft says developers preferred MAI Code 2 over GPT-6 Sol 68% of the time.
MAI Code 2 Flash is a cheaper model for high-volume work such as explaining code and small fixes. Microsoft says it is comparable to GPT-6 Luna at 65% of the price and to Claude Haiku 5.5 at 20%.
Both models are available in GitHub Copilot now. Flash is also in a limited-access preview in Microsoft Foundry, the company's platform for developers building on AI models. Unite.AI reports that Flash is planned for OpenRouter, Vercel, OpenCode and Kilo.
These are Microsoft's own benchmark results. Our benchmark explainer covers why company-run scores need outside checking.
The other five models
- MAI Cyber 2 Flash hunts for security flaws across whole codebases. Microsoft says it matches GPT-6 Astra on the CyberGym-E2E test at under 2% of the cost per task.
- MAI Image 2.7 and a Flash version edit and generate images; Microsoft puts the Flash version at about $10 per 1,000 images. Both are available in Foundry.
- MAI Realtime 1 is Microsoft's first live speech-to-speech model, starting at $0.01 per audio minute.
- Dragon Copilot Transcribe, built on MAI Transcribe 2, now transcribes doctor-patient conversations in Microsoft's Dragon Copilot, which Microsoft says handles about 10 million encounters a month.
- Decision 1.1 routes requests to larger models only when needed. Microsoft says 4,000 customers use it.
Microsoft also said its Maia 200 AI chip is in production, and previewed MAI Licensed Weights, which would let customers customize MAI models and run them in their own Azure environment.
Why it matters
Microsoft offers OpenAI and Anthropic models in GitHub Copilot, and its own announcement measures MAI Code 2 directly against them. Putting its own coding model into GitHub Copilot at a fraction of their prices gives Microsoft a way to cut what it pays outside labs and gives customers a cheaper default for routine coding. The release lands on the same day OpenAI shipped GPT-6.2 Astra and Sol, and follows OpenAI's own price cut with GPT-6.1 Sol. For background on how these prices are quoted, see why AI prices are quoted per million tokens.
Sources
- Seven new world class models at scalable price-performance, Microsoft AI, November 17, 2026
- Microsoft Releases Seven AI Models With Lower-Cost Enterprise Options, Securities.io, November 17, 2026
- Microsoft AI Releases MAI Code 2 and MAI Code 2 Flash for Enterprises, Unite.AI, November 17, 2026
MicrosoftAI modelsOpenAIAnthropic
Written by an AI correspondent under standards set by editor Jacob Knight. Spotted an error? Tell us and we will fix it.