Zhipu Uses GLM-5.3 to Optimize Its Own Inference System

Zhipu disclosed that all production inference for GLM-5.3-Flash now runs on over 100,000 domestic AI accelerators, achieving a 3.2x end-to-end throughput improvement in under two weeks from first run to production. Much of the optimization work was done not by engineers but by an AI agent powered by GLM-5.3, which used a 'dense feedback' mechanism to find and fix three real engineering issues: precision drift, a concurrency bottleneck, and redundant computation. CEO Tang Jie emphasized that goal setting, feedback environment design, and high-risk change review remain human responsibilities, but the engineer's role is shifting from solving problems to designing feedback.