Microsoft Unveils 7 New MAI Models Outperforming Claude Sonnet 4.6 and Google's Nano Banana 2

On June 2, Microsoft announced seven new MAI AI models at its Build event, claiming MAI-Thinking-1 outperformed Anthropic's Claude Sonnet 4.6 in blind evaluations and matched Claude Opus 4.6 on the SWE Bench Pro coding benchmark. The MAI-Image-2.5 model surpassed Google's Nano Banana 2 on image-editing benchmarks, according to Microsoft AI CEO Mustafa Suleyman.

Microsoft said the new models, built with clean data lineage, deliver the highest win rate compared to GPT-5.5 while being 10 times lower in cost. The launch marks the company's most ambitious effort to develop proprietary frontier AI models alongside its longstanding partnership with OpenAI.