Mistral AI has launched Medium 3.1, the latest iteration of its mid-sized large language model. According to the company, the model now ranks second on the widely followed Agent leaderboard, signalling a gain in reasoning and task-completion capabilities while maintaining what it described as a “competitive” price point for developers. The upgrade is already being adopted by third-party platforms. CAMEL-AI, an open-source multi-agent framework, said it has integrated Medium 3.1, enabling users to design, coordinate and deploy AI agents with greater reliability and adaptability. The early uptake suggests the model could see rapid diffusion among developers looking for an alternative to larger-scale systems.