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 product updates and launches from the past 90 days.
Zhipu AIMay 21Z.ai successfully deployed its ZCube network architecture in production to power GLM-5.1 coding services, reducing hardware costs by 33% while boosting throughput. By flattening the network topology, the system eliminates the congestion typically caused by moving massive amounts of data between GPUs during long-context inference.
Zhipu AIMay 8Zhipu AI released a technical report for GLM-5V-Turbo, detailing the architecture and training methods behind its multimodal agent capabilities. The report highlights how native vision integration and scaled reinforcement learning enable the model to perceive and act across complex GUIs and coding tasks.
Zhipu AIApr 30Z.ai identified and fixed low-level race conditions in its GLM-5 inference stack that caused garbled outputs and repetition during high-concurrency coding tasks. By introducing a layer-wise cache storage scheme called LayerSplit, the lab also increased system throughput by up to 132% for long-context workloads.
Zhipu AIApr 27Z.ai extended its high-volume usage promotion for the GLM-5.1 and GLM-5-Turbo models until June 30, 2026. This extension allows developers to run intensive agentic engineering tasks with three times the standard capacity during non-peak hours.
Zhipu AIApr 17Zhipu AI released an updated chat template for the GLM-5.1 model series to fix a critical bug in tool calling. The previous template failed to parse array-formatted tool outputs from frameworks like vLLM, causing models to ignore results and repeat calls indefinitely.
Zhipu AIApr 7Zhipu AI released GLM-5.1, an open-source model that ranks first among open-weights systems on major coding benchmarks like SWE-Bench Pro. The model is designed for long-horizon tasks, capable of running autonomously for eight hours to refine strategies and execute thousands of tool calls.
Zhipu AIApr 1Zhipu AI released GLM-5V-Turbo, a multimodal foundation model specifically architected for vision-based coding and GUI agent workflows. It natively understands images, videos, and design drafts to automate frontend recreation and visual debugging without degrading text-based reasoning.
Zhipu AIMar 31Z.ai released AutoClaw, a local setup tool for the OpenClaw assistant framework that requires no API keys and keeps all data on-device. By pairing the tool with the new GLM-5-Turbo model, users can run complex multi-step agentic workflows entirely within their own infrastructure.
Frequently asked questions