4.7z May 2026

The model has demonstrated high benchmark scores, including 85.7% on GPQA-Diamond and 42.8% on Humanity's Last Exam (HLE) .

These features allow the model to maintain reasoning chains across multiple conversational turns, which is critical for complex tasks rather than resetting the context after every action. The model has demonstrated high benchmark scores, including

GLM-4.7 is accessible via the BigModel.cn API and integrated into various development tools such as OpenRouter , Vercel, and Cursor . Pricing & Access and Cursor . Pricing & Access