Microsoft's June 2026 Build announcement of the MAI model family, headlined by MAI-Thinking-1, established the primary catalyst for trader sentiment on release timing and attribution. The medium-sized sparse MoE reasoning model (35B active parameters) was developed in-house without third-party distillation, delivering competitive results on AIME math benchmarks and SWE-Bench Pro coding tasks while earning higher blind human preference ratings than Claude Sonnet 4.6 in Microsoft evaluations. It advanced to public preview via Microsoft Foundry in August 2026, with ongoing integrations into Copilot, GitHub tools, and Azure Maia silicon. Traders are watching for expanded availability, new capability benchmarks, or follow-on MAI releases that could clarify resolution criteria amid rapid competitive shifts in the reasoning model space.