GLM-5.3: benchmark results

Provider: Zhipu. Released 2026-08-14. Access: API.

Unified ELO 1715 ± 1, rank #36 of 1392 rated models, from 134 benchmark results.

Strongest benchmark results

BenchmarkScoreMetricPercentile
Featherbench100Pass Rate (%)100
Humanity's Last Exam (Self-Reported, With Tools)62.5Accuracy (%)100
LLM Stats (PostTrainBench)39.8Score (%)100
LLM Stats (SWE-Marathon)42.5Score (%)100
SWE-Milestone58.75Milestone Score (%)100
SecIT Bench (Pydantic AI)81.44Accuracy (%)100
Z.ai GLM-5.3 Launch - GDPval-AA v21769ELO100
LiveBench Theory of Mind88.46Score99.1
EQ-Bench Longform Writing81.8Writing Score (0-100)98.1
LLM Stats Score53.6LLM Stats Score (conservative rating)97.7
OpenRouter Tau2-Bench Airline80Accuracy (%)97.5
BenchmarkList ECI148.78Capability Index (ECI)95.1

Interactive version: theaggregate.ai/model?slug=glm-5-3 · How It Works · Data refreshed daily, snapshot 2026-09-05.