The open-weight model GLM 5.2 recently outperformed Claude Opus 4.8 in benchmark evaluations. It currently stands as the highest-performing open-weight option when models are given nothing but a standard prompt.