Shanghai-based AI startup MiniMax announced the release and open-sourcing of its next-generation MiniMax-01 series models. This release includes the foundational language model MiniMax-Text-01 and the visual multimodal model MiniMax-VL-01. The MiniMax-01 series introduces an innovative hybrid structure. This unique approach enables near-linear complexity, allowing the model to efficiently process ultra-long inputs and outputs.