GLM-5 (Non-reasoning): benchmark results
GLM-5 evaluated with reasoning disabled. Provider: Zhipu. Released 2026-02-12. Access: Open.
Aggregate Rating 1620 ± 10, rank #646 of 2168 rated models, from 165 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| EnvSimBench - Difficult | 28 | Config Match (%): share of the 50 Difficult state-changing s | 100 |
| HalluWorld - Grid Uncertainty | 0 | Hallucination rate (%) on the uncertainty probes (abstaining | 100 |
| Tau2-Bench Telecom | 97.37 | Success Rate (%) | 97.6 |
| AA TAU-2 Bench | 97.37 | Accuracy (%) | 97.5 |
| CDR-Bench - Order-Agnostic Recipes | 24.8 | Recipe success RS@3 (%) on the order-agnostic mapper-recipe | 90 |
| UGI - Natural Intelligence | 49.38 | NatInt Score | 88.3 |
| AA Omniscience - Software Engineering (SWE) - PHP | 42 | Accuracy (%) | 87.2 |
| AA Omniscience - Software Engineering (SWE) - Python | 35.5 | Accuracy (%) | 87 |
| AA Omniscience - Software Engineering (SWE) - C | 55 | Accuracy (%) | 86.7 |
| AA Terminal-Bench Hard | 39.39 | Accuracy (%) | 86.5 |
| UGI - Writing | 49.22 | Writing Score | 86.3 |
| AA Omniscience - Software Engineering (SWE) - Go | 32 | Accuracy (%) | 85.5 |
Interactive version: theaggregate.ai/model?slug=glm-5-non-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-20.