On April 8, Zhipu introduced its latest flagship model, GLM-5.1, which was made available on Huawei Cloud immediately upon release, seamlessly integrating with a variety of Huawei Cloud products. This model marks significant advancements in long-duration task management, allowing for continuous autonomous operation for up to eight hours per task while delivering comprehensive, engineering-level results.

GLM-5.1 achieves Layer-level Mixture of Experts (MOE) balancing on Ascend computing power, taking advantage of the unique characteristics of Ascend Attention operators. Through coordinated optimization of the inference framework and hardware, it substantially improves the balance of computing power and HBM memory access. Huawei Cloud has implemented system-wide optimizations, leading to a 30% increase in inference speed and enhanced overall throughput.

Currently, the Huawei Cloud Model as a Service (MaaS) platform offers a hassle-free, one-click deployment for the GLM-5.1 API service, supporting online interaction. Enterprises can deploy inference services with just a single click via the Huawei Cloud ModelArts platform, accommodating both public and dedicated resource pool setups.

Huawei Cloud CodeArts, a code intelligence platform, has integrated GLM-5.1, boosting its capacity to manage complex, real-world engineering tasks and providing free access to users. Additionally, with the backing of GLM-5.1, Huawei Cloud’s AgentArts intelligent agent development platform has seen a notable improvement in tool invocation accuracy and task execution efficiency. This enables the efficient creation of intelligent agents and multi-agent collaboration systems tailored for complex scenarios.

Users also have the option to deploy OpenClaw on Huawei Cloud Flexus to leverage GLM-5.1, improving multi-round task consistency and minimizing daily usage failure rates.