MiniMax introduces the MiniMax M2.7, initiating the self-evolution of its models.

MiniMax introduces the MiniMax M2.7 model, marking the first time a model has deeply participated in self-iteration. The model can construct complex Agent Harnesses, leveraging Agent Teams and complex Skills to complete highly complex productivity tasks. In the software engineering field, M2.7 scored 56.22% in the SWE-Pro benchmark, approaching Opus levels, and achieved 55.6% in VIBE-Pro end-to-end project delivery. In the professional office domain, its GDPval-AA ELO score of 1495 is the highest among open-source models. (Source: Xiyu Technology)