Shanghai AI Unicorn MiniMax officially open source inference model MiniMax-M1. According to MiniMax, this is the world's first large-scale mixed-attention reasoning model with open weights. With a hybrid gated expert architecture combined with Lightning Attention, the M1 achieved significant breakthroughs in performance and inference efficiency. According to experimental data, the M1 series surpassed most closed source models in productivity scenarios such as long context understanding and code generation, and only lagged behind top closed source systems by a slight gap.

Shanghai AI Unicorn MiniMax officially open source inference model MiniMax-M1. According to MiniMax, this is the world's first large-scale mixed-attention reasoning model with open weights. With a hybrid gated expert architecture combined with Lightning Attention, the M1 achieved significant breakthroughs in performance and inference efficiency. According to experimental data, the M1 series surpassed most closed source models in productivity scenarios such as long context understanding and code generation, and only lagged behind top closed source systems by a slight gap.