Microsoft has a reasoning AI model of its own now, and it wants you to know it built the thing itself.
Announced at Build 2026, MAI-Thinking-1 is billed as Microsoft's new flagship in-house model. The company says it trained the model "from the ground up on clean data, without distillation from third-party models" — a pointed claim that distances it from OpenAI, whose models Microsoft has licensed for years. Microsoft describes MAI-Thinking-1 as "medium-sized" and says it "matches leading models" on "key" software engineering benchmarks, though the company did not specify which benchmarks or which rivals it's measuring against.
The timing matters. Microsoft and OpenAI recently renegotiated their partnership to loosen the ties between them, and MAI-Thinking-1 is the clearest signal yet that Microsoft is building toward AI independence rather than permanent reliance on a single supplier. A homegrown reasoning model gives Microsoft more leverage in that relationship and more control over the capabilities it ships into products like GitHub Copilot and Azure.
The "matches leading models" framing is doing a lot of work here. Without named benchmarks and named competitors, it's a marketing claim dressed up as a technical one — a pattern that has become standard practice across every major AI lab. Scrutiny should come when independent evaluators get their hands on it.