Xiaomi is advancing AI technology with its new agent models from the MiMo team, including the flagship MiMo-V2-Pro, which boasts over one trillion parameters and improved processing efficiency. This model performs remarkably well in various benchmarks, ranking high globally while offering competitive pricing at $1 per million input tokens. Xiaomi’s innovative MiMo-V2-Omni integrates multiple sensory inputs for real-time decision-making and web browsing tasks, outperforming competitors in certain audio and image metrics but lagging in others. Additionally, MiMo-V2-TTS excels in speech synthesis, enabling emotional and dynamic voice outputs based on natural language descriptions. Despite impressive initial performance, Xiaomi acknowledges that it needs to enhance features like long-term planning and coordinated multi-agent systems to tackle tasks comparable to those handled by leading models like Claude Opus 4.6. The company’s strategic focus on affordable pricing and comprehensive AI capabilities positions it firmly in the competitive AI landscape.
Source link
