MiniMax Open-Sources M2.7 Model Achieving SOTA Performance on Multiple Benchmarks

MiniMax announced that its M2.7 model is officially open-sourced, achieving state-of-the-art performance with a score of 56.22% on the SWE-Pro benchmark and demonstrating competitive results across multiple evaluation metrics. The model release expands the ecosystem of high-performance open-source language models available through Hugging Face, providing developers with additional options for fine-tuning and deployment. MiniMax's decision to open-source the M2.7 model reflects the continuing trend toward making advanced AI capabilities accessible to the broader developer community through permissive licensing.