Qwen2.5-7B-Instruct-Uncensored: benchmark results
Orion-zhen's uncensored SFT+DPO fine-tune of Qwen2.5 7B Instruct, trained to remove refusals while keeping base capabilities. Provider: Alibaba. Released 2024-09-26. Access: Open.
Unified ELO 1541 ± 1, rank #476 of 1392 rated models, from 130 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Korean LLM Leaderboard | 49.41 | Average Score (%) | 97.5 |
| Open LLM Leaderboard - MATH Level 5 | 47.73 | Score | 95.3 |
| Open Arabic LLM - Arabic MMLU Arabic Language (Grammar) | 54.25 | Accuracy (%) | 85.8 |
| Open LLM Leaderboard - IFEval | 72.04 | Score | 85.7 |
| Open Arabic LLM - Madinah QA Arabic Language (Grammar) | 59.18 | Accuracy (%) | 84.6 |
| Open Arabic LLM - Madinah QA Average | 64.23 | Accuracy (%) | 83.3 |
| Open LLM Leaderboard - MMLU-Pro | 38.07 | Score | 82.1 |
| Open Arabic LLM - Aratrust Unfairness | 92.73 | Accuracy (%) | 82 |
| Open Arabic LLM - Arabic MMLU HT College Computer Science | 49 | Accuracy (%) | 76.5 |
| Open LLM Leaderboard - MuSR | 13.58 | Score | 74.6 |
| Open Arabic LLM - Aratrust Illegal | 90.57 | Accuracy (%) | 73.9 |
| Open Arabic LLM - Arabic MMLU HT Moral Scenarios | 34.41 | Accuracy (%) | 72.8 |
Interactive version: theaggregate.ai/model?slug=qwen2-5-7b-instruct-uncensored · How It Works · Data refreshed daily, snapshot 2026-09-05.