Xiaomi’s late-September release of the MiMo-V2.6-Pro large language model has driven overwhelming trader consensus that the company ranks as the third-best AI lab. The sparse mixture-of-experts model, trained with scaled reinforcement learning on hundreds of thousands of trajectories, posted a 46 score on the Artificial Analysis Intelligence Index—topping all open-weights entries and closing the gap with leading closed systems from OpenAI and Anthropic. Its 1-million-token context, native omnimodal inputs, and record-low inference cost further bolstered perceptions of rapid capability gains. A reversal before month-end would require credible new releases or benchmark revisions from other labs that materially surpass MiMo-V2.6-Pro on independent evaluations.