Signal Microsoft unveils MAI-Code-1-Flash to lessen reliance on OpenAI

Summary

Microsoft unveiled its own coding AI model, MAI-Code-1-Flash, at the Build 2026 conference in San Francisco on June 2, 2026. It is one of seven new MAI models released by Mustafa Suleyman's AI Superintelligence Team, has five billion parameters, and was trained directly on GitHub Copilot's production workflows. The company said the model outperforms Claude Haiku 4.5 by 16 percentage points on SWE-Bench Pro while using 60 percent fewer tokens on complex tasks. Under Copilot's AI Credits pricing it costs $0.75 per million input tokens, $0.075 when cached, and $4.50 per million output tokens, and it is offered on the Free, Student, Pro, Pro+ and Max personal plans, though Business and Enterprise access has no timeline yet. The launch comes after Microsoft's seven-year exclusive partnership with OpenAI ended in April 2026, with chief executive Satya Nadella describing it as a shift from consuming a frontier model to fully participating at the frontier. Independent analysis and the model's own data card noted that OpenAI models were used in preparing its training data, which complicates the claim that it was built fully in-house.