Qwentile2.5-32B-Instruct: benchmark results
maldv's community merge of eight Qwen2.5/QwQ 32B models, aiming to fold QwQ-style reasoning into a steerable instruct model. Provider: maldv. Released 2024-12-19. Access: Open.
Aggregate Rating 1636 ± 24, rank #339 of 1656 rated models, from 184 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open LLM Leaderboard - MMLU-Pro | 58.79 | Score | 99.6 |
| Open LLM Leaderboard - BBH | 69.63 | Score | 98.7 |
| Open Japanese LLM - MR | 94.6 | Score (%) | 98.4 |
| Open Portuguese LLM - ENEM | 85.16 | Accuracy (%) | 98.2 |
| Open LLM Leaderboard - MATH Level 5 | 52.19 | Score | 97.9 |
| Open Portuguese LLM - BLUEX | 78.44 | Accuracy (%) | 97.6 |
| Open Japanese LLM - HE | 78.64 | Score (%) | 97.4 |
| Open Japanese LLM - Jmmlu Exact Match | 76.39 | Score (%) | 97.4 |
| Open Arabic LLM - Aratrust MentalHealth | 96.05 | Accuracy (%) | 97.2 |
| Open Arabic LLM - Alghafa Multiple Choice Grounded Statement Soqal Task | 94.67 | Accuracy (%) | 96.9 |
| Open Japanese LLM - Jsem Exact Match | 80.93 | Score (%) | 96.9 |
| Open Portuguese LLM - OAB Exams | 65.51 | Accuracy (%) | 96.9 |
Interactive version: theaggregate.ai/model?slug=qwentile2-5-32b-instruct · How It Works · Data refreshed daily, snapshot 2026-10-20.