Microsoft AI's MAI-Thinking-1: Getting Models to Think Is Easy, Sustained Thinking Is Hard

Microsoft AI 的 MAI-Thinking-1：让模型思考不难，让它持续思考才难

Microsoft AI says MAI-Thinking-1 uses three mechanisms—thermostat, circuit breaker, and self-distillation—to keep RL training stable for several thousand steps; the RSS snippet contrasts MAI’s discipline with DeepSeek’s efficiency and GLM’s endurance.

Why it matters: HKR-H/K/R all pass: the hook is training persistence, the new facts are three stability mechanisms and thousand-step RL runs, and the audience cares about reasoning-model stability. Not a major model launch, so it stays below 85.