< Back to situation

[REVISION HISTORY]

Zhipu AI release of GLM-5.3 model

Updated 1 time since CLSTR started tracking revisions of this situation.

What changed

2026-08-16 15:16 UTC → 2026-08-17 01:39 UTC · added removed

Zhipu AI release of GLM-5. 3 GLM-5.3 model

Zhipu AI (also referred to as Z. ai) has released GLM-5. 3, GLM-5.3, a large language model optimized for programming and cybersecurity. Although the model utilizes the same 743-billion-parameter Mixture of Experts base as GLM-5. 2, GLM-5.2, it achieves performance gains through expanded post-training scaling. In technical benchmarks, the model shows a 50% improvement in coding tasks and an 84.5% success rate in CyberGym cybersecurity testing, surpassing competitors such as Anthropic’s Mythos 5 and OpenAI’s GPT-5. 6 Sol. The model features a 1 million token context window and is designed for long-horizon tasks and agentic capabilities, including tool calling via MCP. GLM-5. 3 While GLM-5.3 shows strength in vulnerability detection, outperforming Mythos 5’s 83.8% with its 84.5% score, it trails Western competitors in offensive capabilities. In the Exploit Bench test, which measures the ability to convert vulnerabilities into functional attacks, GLM-5.3 scored 54.4% compared to Mythos 5’s 78.0%. GLM-5.3 is currently accessible via API, the GLM Coding Plan, and ZCode. The company intends to release the full model weights to the public following safety evaluations and hardening processes.

Versions

  1. 2026-08-17 01:39 UTC Zhipu AI release of GLM-5.3 model
  2. 2026-08-16 15:16 UTC Zhipu AI release of GLM-5. 3 model

Only revisions since CLSTR began indexing content versions appear here. Select a version to see what changed compared to the one before it.