glm-4-9B (0414): benchmark results
Zhipu's April 2025 GLM-4-9B-0414 chat release, optimized for large-scale batch tasks such as translation; distinct from the earlier GLM-4-9B base and chat releases. Provider: Zhipu. Released 2025-04-14. Access: Open.
Aggregate Rating 1502 ± 30, rank #2196 of 10870 rated models, from 38 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Portuguese LLM - FaQuAD NLI | 86.86 | Macro F1 (%) | 99.6 |
| Open Portuguese LLM - ASSIN2 RTE | 93.79 | Macro F1 (%) | 92 |
| Open Portuguese LLM - ENEM | 74.32 | Accuracy (%) | 86.7 |
| FinSafetyBench (PAIR) - Ethical Violations (English) | 71.31 | Attack success rate (%; PAIR jailbreak; share of the benchma | 83.3 |
| Open Portuguese LLM - BLUEX | 63.14 | Accuracy (%) | 83.1 |
| Open Portuguese LLM - OAB Exams | 49.66 | Accuracy (%) | 73.9 |
| FinSafetyBench (FlipAttack) - Ethical Violations (English) | 20.02 | Attack success rate (%; FlipAttack jailbreak; share of the b | 66.7 |
| FinSafetyBench (FlipAttack) - Financial Crimes (English) | 34.51 | Attack success rate (%; FlipAttack jailbreak; share of the b | 66.7 |
| FinSafetyBench (PAIR) - Financial Crimes (English) | 90.52 | Attack success rate (%; PAIR jailbreak; share of the benchma | 66.7 |
| CommunityBench - Community-Consistent Generation | 176.85 | BTL-Elo rating from majority-voted LLM-judge pairwise compar | 56.2 |
| EHRBench - Diagnosis | 58.36 | Accuracy (%) on the diagnosis-decision questions (complete a | 53.3 |
| EHRBench - MIMIC-III | 61.92 | Accuracy (%) on the questions built from MIMIC-III records, | 50 |
Interactive version: theaggregate.ai/model?slug=glm-4-9b-0414 · How It Works · Data refreshed daily, snapshot 2026-11-12.