On June 17, MiniMax, a domestic AI unicorn headquartered in Shanghai, unveiled MiniMax-M1, the world's pioneering open-source large-scale hybrid architecture inference model, on its official website and GitHub platform. This breakthrough swiftly propelled MiniMax-M1 to the second spot on the prestigious global ranking of open-source models, trailing only behind DeepSeek-R1-0528, which was introduced on May 28. Yan Junjie, the founder and CEO of MiniMax, shared his sentiments on WeChat Moments, noting that for the first time, he perceived mountains as scalable. Within a mere four business days of M1's release, MiniMax demonstrated its momentum by successively launching Hailuo 02, a video generation model, MiniMax Agent, a general intelligent agent, Hailuo Video Agent, a video creation intelligent agent, and Voice Design, a voice design tool. The company maintained a daily update schedule, capturing widespread attention. M1 distinguished itself in authoritative evaluations, particularly showcasing substantial advantages in long text processing and tool invocation.