On June 17, on the official website and open source platform GitHub, MiniMax, a domestic AI unicorn from Shanghai, launched the world's first open source large-scale hybrid architecture inference model, MiniMax-M1. The model is already ranked second in the global open source model in authoritative reviews. On the day M1 was released, MiniMax founder and CEO Yan Junjie wrote in his circle of friends: For the first time, I felt that the mountains were not insurmountable. In the 4 working days after M1, MiniMax also continuously released the video generation model Hailuo 02, the general intelligent MiniMax Agent, the video creation agent Hailuo Video Agent, and the sound design tool Voice Design, keeping things up and loud. As soon as M1 debuted, it ranked second in the global open source model in the authoritative evaluation list, after DeepSEEK-R1-0528, which was released on May 28. However, judging from the segments, M1 has a crushing advantage in long text processing, tool calls, etc.

On June 17, on the official website and open source platform GitHub, MiniMax, a domestic AI unicorn from Shanghai, launched the world's first open source large-scale hybrid architecture inference model, MiniMax-M1. The model is already ranked second in the global open source model in authoritative reviews. On the day M1 was released, MiniMax founder and CEO Yan Junjie wrote in his circle of friends: For the first time, I felt that the mountains were not insurmountable. In the 4 working days after M1, MiniMax also continuously released the video generation model Hailuo 02, the general intelligent MiniMax Agent, the video creation agent Hailuo Video Agent, and the sound design tool Voice Design, keeping things up and loud. As soon as M1 debuted, it ranked second in the global open source model in the authoritative evaluation list, after DeepSEEK-R1-0528, which was released on May 28. However, judging from the segments, M1 has a crushing advantage in long text processing, tool calls, etc.