Microsoft said at its Ignite conference that it rolled out seven specialized MAI models, replacing OpenAI and Anthropic systems across its products, including its coding tools and cybersecurity defenses.
The flagship, MAI Code 2, now handles deep debugging, multi-file edits and repository-wide reasoning on complex coding tasks. Microsoft said it passes 63.2% of coding tasks on its tests, versus 64.1% for OpenAI's GPT-6.1 Sol and 65.3% for Anthropic's Claude Opus 5.5.
It costs $0.309 per task, against $0.761 for GPT-6.1 Sol and $6.306 for Claude Opus 5.5, Microsoft said. The pricing holds at a promotional rate through May 17, 2027.
A second model, MAI Cyber 2 Flash, is built to defend against cyberattacks and scored 92.3% on the CyberGym-E2E benchmark against 91.5% for OpenAI's GPT-6 Astra, at $0.45 per task versus $34.38, Microsoft said. The company plans to start automatically swapping it in for outside models across its security tools in mid-December.
Pareekh Jain, chief executive of EIIRTrend and Pareekh Consulting, said the move cuts Microsoft's reliance on OpenAI and Anthropic while leaning on Azure, Visual Studio and GitHub to ‘lock developers into its own tools.’
Greyhound Research's chief analyst, Sanchit Vir Gogia, cautioned that ‘the gap narrows sharply once the promotion ends.’
Microsoft also showed an image model, MAI Image 2.7, a voice model priced at $0.01 a minute of audio, and a routing model it says runs 2.2 times faster and 4.4 times cheaper than typical large language models.


