Skip to content

AI · 17 November 2026 · 3 min read

Microsoft releases MAI Code 2 and a family of seven models

Microsoft said on 17 November, at its Ignite conference, that it is releasing seven models. On its own average of four coding tests, it says MAI Code 2 reaches 98 percent of GPT-6.1 Sol at 40 percent of the price, and 96 percent of Claude Opus 5.5 at 5 percent of the price. It says the models are not yet tuned for fundamental science or very long agent work.

A chart graphic with the words MAI Code 2 and MAI Code 2 Flash over faded percentage lines.
Microsoft’s graphic for MAI Code 2 and MAI Code 2 Flash. It is not a photograph of a data centre or a test.

Microsoft said on 17 November that it is releasing a family of seven models. The note names MAI Code 2 for everyday enterprise coding, MAI Code 2 Flash for high-volume routine coding, and MAI Cyber 2 Flash for a cybersecurity benchmark. It says MAI Code 2 can be tried in GitHub Copilot. On an average of four tests it uses for Copilot work, DeepSWE 1.1, Terminal Bench 2.1, SWE-Atlas and FrontierCode v1.1, Microsoft says MAI Code 2 gives 98 percent of the performance of GPT-6.1 Sol at 40 percent of the price, and 96 percent of Claude Opus 5.5 at 5 percent of the price.

Microsoft says MAI Code 2 Flash offers comparable performance to GPT-6 Luna at 65 percent of the price, and to Claude Haiku 5.5 at 20 percent of the price, on three of those coding tests. It says MAI Cyber 2 Flash matches GPT-6 Astra’s quality at less than 2 percent of the cost per task on a benchmark it calls CyberGymE2E. The same note says the models are built for coding customers do every day, and are explicitly not yet tuned for harder problems, including fundamental science research and very long-horizon agent work.