June 2026 — Chinese AI developer Z AI has introduced GLM-5.2, which has emerged as the leading open-weights model on the latest Artificial Analysis Intelligence Index v4.1. The model achieved an Intelligence Index score of 51, placing it ahead of all other publicly available open-weights competitors and reinforcing the growing competitiveness of the open AI ecosystem.
According to benchmark results published by Artificial Analysis, GLM-5.2 currently occupies the top position among open-weights models, surpassing notable rivals including MiniMax‘s MiniMax-M3, DeepSeek‘s DeepSeek V4 Pro, and Moonshot AI‘s Kimi K2.6.
Strong Performance Through Advanced Reasoning
A key factor behind GLM-5.2’s performance is its extensive use of reasoning during inference. Benchmark evaluations indicate that the model generates approximately 43,000 output tokens per task, of which roughly 37,000 tokens are dedicated to reasoning processes. This approach contributes to stronger benchmark results, particularly on complex reasoning and agentic tasks, though it also results in higher computational and token costs compared with some competing models.
Significance for the Open-Weights Ecosystem
The achievement marks another milestone for the open-weights AI community. Over the past year, open models have narrowed the performance gap with proprietary frontier systems at an unprecedented pace. GLM-5.2’s ranking demonstrates that open-weights models are increasingly capable of delivering state-of-the-art performance across reasoning, coding, tool use, and autonomous agent workflows.
The result also highlights the growing influence of Chinese AI laboratories in the global AI landscape. Organizations such as Z AI, DeepSeek, Moonshot AI, and MiniMax have become major contributors to advances in open AI models, accelerating innovation and increasing competition across the sector.
Position Relative to Frontier Proprietary Models
While GLM-5.2 now leads the open-weights category, proprietary frontier systems continue to hold the highest overall positions on intelligence benchmarks. Models such as Claude Opus 4.8 remain ahead on aggregate intelligence measures, indicating that a performance gap still exists between the best open and closed systems.
Nevertheless, the gap has narrowed considerably, and GLM-5.2’s performance underscores the rapid pace at which open-weights models are approaching frontier-level capabilities.
