←── back to feed
/topics/glm-5-3-frontier-coding-model-release

GLM-5.3 frontier coding model release

6 items3 sourcesupdated 36d agotrend 0

Z.ai released GLM-5.3 on August 14, 2026, a frontier coding model built by scaling post-training on the unchanged 743B GLM-5.2 base. The model shows substantial gains on code and long-horizon reasoning tasks, with Terminal-Bench 3.0 jumping from 4.6 to 28.3 and cybersecurity benchmarks more than doubling.

  • Terminal-Bench 3.0: 4.6 → 28.3; DeepSWE v1.1: 46.2 → 66.9
  • CyberGym reached 84.5%; ExploitBench more than doubled to 54.4%
  • Reuses 743B GLM-5.2 base model without retraining
  • Weights available in approximately two weeks
  • Gains driven entirely by scaled post-training across more task environments and longer training