GLM-5.2 Becomes Top Open Weights Model on Artificial Analysis Intelligence Index

GLM-5.2 scores 51 on the Artificial Analysis Intelligence Index, leading over MiniMax-M3 and DeepSeek V4 Pro, both scoring 44.

The model retains the same size as its predecessor, GLM-5.1, with 744 billion total parameters and 40 billion active parameters, but shows notable improvements in evaluations, especially in scientific reasoning and other metrics.

GLM-5.2 scores 1524 on GDPval-AA v2, outperforming MiniMax-M3 and DeepSeek V4 Pro, and is comparable to proprietary models like GPT-5.5.

It uses 43,000 output tokens per task, which is an increase from GLM-5.1 and higher than other open weights models, indicating a lower token efficiency at its intelligence level.

The model is positioned on the Pareto frontier of Intelligence vs Cost per Task, costing approximately $0.46 per task, which is higher than some competitors but reflects its leading intelligence score.

GLM-5.2 is licensed under MIT and features a context window of 1 million tokens, significantly larger than the previous version. It is available through Z ai's first-party API and various third-party providers.