The latest AI news and updates of Zhipu AI — AI lab behind the GLM model series and the Z.ai developer platform. Covering Zhipu AI's latest launches, product updates, and company news from the past 90 days.
Zhipu AIOct 12Z.ai launched GLM-Legal, a domain-specific model built on GLM-5.3-Flash. It matches frontier model performance on mainstream legal benchmarks and handles complex tasks like contract review, evidence analysis, and compliance. The model is available for API access by application and integrates into the AutoClaw agent platform for production legal workflows.
Zhipu AISep 17Z.ai built the inference infrastructure for GLM-5.3-Flash using an Infra Agent powered by GLM-5.3. The system reached production readiness in under two weeks and tripled end-to-end throughput. The agent utilized dense feedback—including local correctness tests, execution traces, and microbenchmarks—to perform targeted hypothesis testing rather than relying on aggregate performance metrics alone.
Zhipu AIAug 28Z.ai released the open weights for GLM-5.3, its most capable model for agentic coding and cyber defense. The model is now available to download, run, and customize on Hugging Face. This release follows the completion of safety evaluation and hardening, fulfilling the commitment to publish the weights two weeks after the initial launch.
Zhipu AIAug 27Z.ai will release the open weights for its GLM-5.3 model on Hugging Face tomorrow, August 28, 2026. The model provides frontier coding performance and emergent cyber defense capabilities, including vulnerability discovery. This release makes the model weights publicly available for local deployment and self-hosting.
Zhipu AIAug 14Z.ai launched GLM-5.3, an open model post-trained on a 743B parameter base model. It delivers improved coding and agentic performance alongside emergent cyber defense capabilities, including vulnerability discovery. The model is available now through the GLM Coding Plan and ZCode, with open weights scheduled for release in two weeks.
Zhipu AIAug 11Z.ai reached 1 million users for its ZCode agentic coding harness. The platform reset usage limits for all GLM Coding Plan subscribers as a community thank-you. A new update improves engineering workflow intelligence and achieves a 98% cache hit rate, delivering approximately 1.8 times more usage capacity for the same computational resources.
Frequently asked questions