EZO-Qwen2.5-32B-Instruct: benchmark results
Axcxept's Japanese-focused instruction tune of Qwen2.5 32B Instruct, trained on curated Japanese Wikipedia and FineWeb data. Provider: AXCXEPT. Released 2024-09-21. Access: Open.
Aggregate Rating 1592 ± 26, rank #471 of 1656 rated models, from 182 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Portuguese LLM - ASSIN2 RTE | 94.89 | Macro F1 (%) | 99.3 |
| Open Arabic LLM - Arabic MMLU HT Moral Scenarios | 58.55 | Accuracy (%) | 98.8 |
| Open Portuguese LLM - BLUEX | 78.3 | Accuracy (%) | 97.4 |
| Open Japanese LLM - Jsts Spearman | 88.23 | Score (%) | 96.6 |
| Open Portuguese LLM - ENEM | 83.14 | Accuracy (%) | 96.3 |
| Open Portuguese LLM - FaQuAD NLI | 82.87 | Macro F1 (%) | 95.6 |
| Open Japanese LLM - Jcommonsenseqa Exact Match | 96.25 | Score (%) | 93.6 |
| Open Arabic LLM - Aratrust Offensive | 94.2 | Accuracy (%) | 93.5 |
| Open Portuguese LLM - OAB Exams | 61.37 | Accuracy (%) | 93.5 |
| Open Japanese LLM - Jsick Exact Match | 85.51 | Score (%) | 93.4 |
| Open Arabic LLM - Arabic MMLU History (Primary School) | 72.55 | Accuracy (%) | 92 |
| Open Arabic LLM - Arabic MMLU Islamic Studies (High School) | 75.15 | Accuracy (%) | 92 |
Interactive version: theaggregate.ai/model?slug=ezo-qwen2-5-32b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.