MiniMax-M3, developed by MiniMax, has officially launched on the Venice platform, marking it as the first open-weight model designed for advanced coding and agentic performance alongside its native multimodal capabilities. The model aims to handle long-horizon, complex workflows, with performance metrics including 59.0% SWE-Bench Pro and 66.0% Terminal Bench 2.1. Users can interact with MiniMax-M3 anonymously, while the weights and a detailed technical report are set to be made available in approximately ten days.

MiniMax: MiniMax is an AI company operating under the handle @MiniMax_AI and focused on developing advanced frontier models. The company recently introduced MiniMax-M3 as its latest model for coding, agentic workflows, and complex reasoning. The model is now available for interaction on the Venice platform.
MiniMax-M3: MiniMax-M3 is MiniMax’s latest frontier model designed for coding, agentic workflows, and complex reasoning tasks. It is the first open-weight model to combine frontier coding and agentic performance with native multimodality and supports long-horizon complex workflows. The model is now live on Venice and available for anonymous use.

Model Access: Weights and the technical report for MiniMax-M3 are scheduled for release in approximately ten days.
Platform Integration: MiniMax-M3 is now live and accessible for chatting on the Venice platform.