Chinese AI Lab Z.ai Releases Open-Source GLM-5.2 With Million-Token Context Window

Chinese AI lab Z.ai has released GLM-5.2, a 753-billion parameter Mixture-of-Experts model with a 1-million-token context window, now available under an MIT open-source license. The model, which features 40 active parameters and is text-only input, was first made available to coding plan subscribers on June 13th before a public release on June 16th, according to Simon Willison's blog. "Z.ai launches GLM-5.2 with 1-million-token context window; MIT-licensed release coming next week" Hacker News reported that initial access was limited to subscribers of the GLM Coding Plan, while the broader MIT-licensed release is expected next week. This marks a significant leap from the previous GLM-5.1 model, which had a 200,000-token context window. The release is positioned as a direct response to US export restrictions targeting Anthropic models, according to Pandaily. The outlet noted that the move showcases "China's push for AI self-reliance and open-source alternatives amid escalating tech tensions between the US and China." "The key challenge addressed is making long context engineering-usable by ensuring reliability under real engineering pressure, rather than simply accepting more tokens." Hugging Face highlighted that GLM-5.2 is designed specifically for long-horizon tasks, delivering substantial improvements over GLM-5.1 by maintaining quality across long, complex coding-agent trajectories. The model's massive context window could enable processing of extremely long documents or codebases in a single pass, as noted by Hacker News. Simon Willison's blog reported strong community buzz around the release, underscoring the model's potential impact on open-source AI development. With its MIT license and million-token capability, GLM-5.2 represents a notable advancement in making long-context AI models accessible to developers worldwide.

4 sources•19 JunRead the Baker's Take→