Xiaomi opens MiMo-V2.6 for agentic RL tests Xiaomi released and open-sourced MiMo-V2.6, pairing open weights with reinforcement-learning resources for agent, coding, cyber and multimodal tasks.
