GLM 4.5 Air: benchmark results
Zhipu's lightweight GLM-4.5 Air open MoE (106B total, 12B active). Provider: Zhipu. Released 2025-07-28. Access: Open.
Aggregate Rating 1616 ± 7, rank #392 of 1656 rated models, from 776 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Galileo Agent - Investment TSQ | 96 | Task Success Quality (%) | 100 |
| LLM Stats (TAU-bench Airline; GLM optimized user simulator) | 60.8 | Score (%) | 100 |
| Polish Legal RAG - Legal Reasoning | 87.72 | Score (0-100) | 100 |
| SEAL - Fortress | 63.18 | Score | 98.6 |
| Galileo Agent - Banking TSQ | 94 | Task Success Quality (%) | 97.6 |
| Galileo Agent - Insurance TSQ | 94 | Task Success Quality (%) | 97.6 |
| ReLE - Agents and Tool Use | 73 | Accuracy (%) | 97.6 |
| EuroEval Lithuanian NLU - WikiANN LT | 70.41 | Micro F1 excluding MISC (%) | 96.1 |
| EuroEval Latvian NLU - Fullstack NER LV | 71.41 | Micro F1 excluding MISC (%) | 95.7 |
| Galileo Agent - Telecom TSQ | 94 | Task Success Quality (%) | 95.2 |
| EuroEval English Knowledge | 94.32 | Knowledge Average Score (%) | 95 |
| EuroEval Icelandic NLU - MIM-GOLD NER | 75.93 | Micro F1 excluding MISC (%) | 94.4 |
Interactive version: theaggregate.ai/model?slug=glm-4-5-air · How It Works · Data refreshed daily, snapshot 2026-10-20.