Zhipu AI released its open-weight GLM-5.2 model on June 21, 2026. Researchers claimed GLM-5.2 matches Anthropic's Mythos model in specific bug-finding and cybersecurity scenarios.

GLM-5.2 is an open-weight model capable of being downloaded and run on readily available hardware. The model performs behind models from Anthropic and OpenAI on general tasks.

OpenRouter token traffic for GLM-5.2 increased at a faster rate compared to the traffic seen after DeepSeek's V4 launch in April. DeepSeek launched its V4 model in April 2026. Gabe Pereyra, co-founder of Harvey, stated, "I've been consistently surprised by how quickly the open source has caught up."

The White House has worked to restrict China's access to powerful AI models, including Anthropic's Mythos and Fable. It has also worked to restrict China's access to the hardware needed to train and run advanced AI models. Pereyra added, "GLM 5.2, you're seeing the first model where it's really competitive with some of these closed-source frontier models."

Anthropic pulled its Fable Mythos-class model following an order by the Trump administration. The Trump administration views Mythos and other advanced AI models capable of identifying vulnerabilities as serious national security threats. On June 26, 2026, OpenAI unveiled its GPT-5.6 model and announced it was limiting access to it due to a government request.

GLM-5.2 is within a percentage point of Anthropic's Opus 4.8 on one agentic benchmark. GLM-5.2 also costs approximately one-fifth of the amount of Anthropic's Opus 4.8.

Why It Matters

Zhipu AI's release of the GLM-5.2 model takes place during U.S. government efforts to restrict China's access to advanced artificial intelligence technologies. The model's ability to operate on consumer hardware and its comparable performance in specific cybersecurity applications to Western models like Anthropic's Mythos occur while the U.S. government has limited access to certain AI models and related hardware.