Xiaomi launched MiMo-V2.6 Pro and Flash AI models after demonstrating reinforcement learning training publicly. The models lead open-source benchmarks but trail closed models like Claude Opus 5 and GPT-6 Astra. Xiaomi plans to open-source the RL training methodology. Pricing remains aggressive at $0.14–0.28/million tokens (Flash) and $0.435–0.87/million tokens (Pro), positioning MiMo-V2.6 as cost-competitive in the sub-$1 workload category on Artificial Analysis indices.