Z ai has released GLM-5.2, establishing it as the new top-ranked open-weight model on the Artificial Analysis Intelligence Index v4.1 with a score of 51. Maintaining the same architecture as its predecessor, the model retains a 744-billion total and 40-billion active parameter configuration while delivering an 11-point performance jump. This advancement positions GLM-5.2 ahead of key competitors including MiniMax-M3 and DeepSeek V4 Pro Max, both scoring 44. The model demonstrates substantial gains across diverse evaluation suites, with particular strength in scientific reasoning. Benchmarks include a 16-point increase to 21 percent on CritPt, a 12-point rise to 40 percent on the Higher Level Entailment dataset, and a 15-point improvement to 27 percent in tau3 banking simulations. Performance on AA-LCR, SciCode, TerminalBench v2.1, and GPQA Diamond also saw notable upgrades, with the latter reaching 89 percent. Additionally, GLM-5.2 achieves a score of 4 on the Artificial Analysis Omniscience Index, reflecting higher factual accuracy and a reduced hallucination rate compared to its predecessor. In agentic workflows, GLM-5.2 dominates the GDPval-AA v2 benchmark, which measures real-world multi-turn agent performance. Scoring 1524, it surpasses MiniMax-M3 and DeepSeek V4 Pro Max, while remaining competitive with the proprietary GPT-5.5 xhigh reasoning tier. The model handles an expanded one-million-token context window, up from 200,000, enabling longer and more complex task execution. Despite generating 43,000 output tokens per task, an increase from GLM-5.1 and exceeding peer open-weight models, GLM-5.2 maintains a favorable position on the Intelligence versus Cost per Task Pareto frontier. Priced at $1.4 per million input tokens, $0.26 for cache hits, and $4.4 for output tokens, the model costs approximately $0.46 per task. While its token consumption is higher than some competitors, its pricing structure and performance efficiency keep it among the most cost-effective solutions at its intelligence tier. Licensed under the permissive MIT agreement, GLM-5.2 is immediately available through Z ai first-party API and a broad network of third-party infrastructure providers, including DeepInfra, Novita, Nebius, Siliconflow, and Fireworks. The release signals a significant milestone in open-weight capabilities, effectively closing the gap between community-driven models and top-tier proprietary systems in both reasoning and autonomous task execution.