Z.ai’s GLM-5.3 has achieved the highest score among open models, with 60 points on the Artificial Analysis Intelligence Index. This ties it with Kimi K3 for the top position, while outperforming the previous GLM-5.2 by seven points. The model shows significant improvement in agentic tasks, with its Elo score on the GDPval-AA v2 benchmark rising from 1,524 to 1,770, placing it second overall behind Claude Opus 5. According to Artificial Analysis, GLM-5.3 is priced at $0.68 per task, which is 19 percent cheaper than Kimi K3’s $0.84, but 1.5 times the cost of its predecessor, GLM-5.2. The model is available via Z.ai’s API, but the company has delayed the release of the open weights by about two weeks. Z.ai stated that GLM-5.3 is so effective at detecting security vulnerabilities that it is first strengthening controls and restricting full access to select security partners. The delay reflects the company’s cautious approach to ensuring the model’s security and reliability before broader access. The model’s performance and pricing position it as a strong contender in the open-model landscape.
Source: thedecoder
