Key facts
- Z.ai has launched GLM 5.2, a new Chinese AI model.
- GLM 5.2 claims performance on par with Anthropic's Claude Opus 4.8 and OpenAI's GPT 5.5.
- The model boasts a 1 million token context window.
- It is designed for long coding tasks and excels in maintaining quality over extended coding sessions.
- GLM 5.2 is an open-source model with no regional restrictions.
A new artificial intelligence model from China, GLM 5.2, has been released by Z.ai, positioning itself as a competitor to leading US AI models such as Anthropic's Claude Opus 4.8 and OpenAI's GPT 5.5. The launch occurred shortly after the US temporarily banned Anthropic from supplying its models to non-Americans.
Z.ai claims GLM 5.2 demonstrates performance nearly on par with its US counterparts, particularly in handling long coding tasks. The model features a substantial 1 million token context window, enabling it to process and retain a large amount of information, equivalent to approximately 750,000 words. This capability is designed to maintain quality across extended and complex coding-agent trajectories.
In performance benchmarks focused on long, intricate coding projects, GLM 5.2 trails Anthropic's Opus 4.8 by only 1% while surpassing GPT 5.5 and Opus 4.7. It also ranks second to Opus 4.8 in tests measuring its ability to enhance smaller models using a single GPU. On the most demanding marathon-length engineering tasks, such as compiler construction, GLM 5.2 trails Opus 4.8 by 13% but remains the second-best performer overall, according to Z.ai.
GLM 5.2 is also highlighted as an open-source model with no regional limitations, allowing users worldwide to modify and share it freely. This contrasts with the closed-source models offered by companies like Anthropic and OpenAI. The development comes amid an ongoing AI race between the US and China, with the US focusing on restricting semiconductor access and China promoting cheaper, open-source models like GLM 5.2 and DeepSeek's R1.