Microsoft launches MAI Code 2 and MAI Cyber 2 Flash at Ignite 2026

Microsoft announced on Tuesday, November 17, the MAI Code 2 and MAI Code 2 Flash, its new in-house coding models, during Microsoft Ignite 2026 in San Francisco. Both are already available in GitHub Copilot, with a straightforward pitch: performance close to the best models from OpenAI and Anthropic, for a fraction of the cost.

According to Microsoft, MAI Code 2 is a general-purpose model for everyday coding work: deep debugging, changes spanning many files, reasoning across entire repositories, and building interfaces and visualizations such as dashboards and charts. MAI Code 2 Flash is the smaller, faster version, aimed at high-volume jobs such as explaining code, making targeted edits and fixing routine bugs. Both are designed for agentic coding, in which the AI plans and carries out several steps on its own.

The numbers — and the caveat

The performance numbers come from Microsoft itself, based on internal testing, and have not yet been confirmed by independent benchmarks. Averaged across four coding benchmarks (DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1), the company says MAI Code 2 delivers 98% of the performance of GPT-6.1 Sol at 40% of the cost and 96% of Claude Opus 5.5's performance at just 5% of the cost. In blind tests of front-end apps, reviewers preferred MAI Code 2's results over GPT-6 Sol 67% of the time.

MAI Code 2 Flash, meanwhile, is said to offer comparable performance to GPT-6 Luna and Claude Haiku 5.5 at 65% and 20% of the cost, respectively, across three benchmarks of routine tasks. There is one important detail: according to Microsoft, the cost figures use the models' 50% promotional price — so the advantage may narrow once the promotion ends.

Microsoft itself points out the limits: the models are not yet tuned for the hardest problems, such as fundamental science research and very long-horizon agentic work.

MAI Cyber 2 Flash and the decision model

The coding pair was not the only news. Microsoft also introduced MAI Cyber 2 Flash, which it calls its most capable security model, with 12 billion active parameters. It will be offered to a limited group of customers, to deal with the surge in vulnerabilities that AI tools are increasingly bringing to light. According to the company, the model outperforms GPT-6 Astra on a cybersecurity benchmark for less than 2% of the cost per task.

Rounding out the package is Microsoft-Decision-1.1, a fast, low-cost model for the small decisions agents make all the time — routing a request to the right tool, classifying, prioritizing, checking whether a step worked and controlling the workflow — and MAI Work Flash, a version of Code 2 Flash tuned for Microsoft 365 apps such as Word, Excel, PowerPoint and Teams, plus Fabric IQ.

Where to use them

Both coding models show up in the GitHub Copilot model picker, billed on usage. MAI Code 2 is available on the Pro, Pro+, Max, Business and Enterprise plans; Flash also reaches the Free and Student plans. In Microsoft Foundry, MAI Code 2 Flash is in a limited-access preview. Microsoft says the models will also come soon to OpenRouter, Vercel, Hermes, OpenCode and Kilo, and that enterprise customers will be able to license the model weights to customize them with their own data, deployed on Azure.

“This isn't just one model. It's not a flash in the pan,” Microsoft AI chief Mustafa Suleyman told Reuters. “This is a system, a hill-climbing machine that we have built, which co-optimizes performance and price.” The strategy is clear: after years of leaning on OpenAI's models, Microsoft wants to compete with its own models on price — and the launch comes on the same day the company unveiled other AI news at Ignite, such as MAI Image 2.7, its image generation and editing model.

This article is part of Geekiki's coverage of Ignite 2026 — also read the public Copilot plugin catalog and the AI Transparency Initiative, announced by Satya Nadella.

Sources

Image credit: Microsoft AI logo, via Wikimedia Commons (public domain).

Transparency: this article was produced with the help of artificial intelligence, based on cross-checking public posts and sources on the internet. We recommend consulting the original sources listed above to confirm the details.

By Muse