Microsoft Drops MAI-Thinking-1, Its First Homegrown Reasoning Model
Microsoft built an advanced reasoning AI model entirely in-house, no third-party distillation involved.
Microsoft just pulled the curtain back on MAI-Thinking-1 — its first advanced reasoning model built entirely from scratch. The key flex here: trained on clean data with zero distillation from third-party models. No borrowing from OpenAI's playbook or anyone else's.
The model dropped alongside six other new in-house AI models unveiled at Build 2026, signaling Microsoft is serious about reducing its dependency on external AI partners.
Building a reasoning model from the ground up is no small feat. It requires massive compute, proprietary training data pipelines, and a team that knows what it's doing. Microsoft clearly wants the world to know it can go toe-to-toe with the best reasoning models out there — on its own terms.
Seven models in one announcement is a statement. Microsoft isn't just playing the AI game anymore. It's building its own deck.