MiniMax M2.7 Model Receives SGLang Framework Support on Launch Day

ME News Report, April 12 (UTC+8), MiniMax officially announced that its M2.7 model received Day-0 support for the SGLang framework on its launch day. According to the article, M2.7 is MiniMax’s first model to deeply participate in its own evolution, achieving a 30% performance improvement through over 100 rounds of autonomous optimization, and earning a 66.6% medal rate on MLE Bench Lite. The model is specifically designed for real-world productivity tasks, excelling at building complex intelligent agent systems and handling fine-grained productivity work, with outstanding performance in evaluations such as SWE-Pro, GDPval-AA, and Toolathon, and native support for multi-agent collaboration. The SGLang framework provides optimized deployment configurations for multiple GPUs from NVIDIA and AMD, including tensor parallelism, expert parallelism settings, and key parameters like enabling function calls and thinking modes. After deployment, it can be accessed via OpenAI-compatible API calls. The model is licensed under the Modified-MIT license. (Source: InFoQ)