AceGPT-v2-32B: benchmark results
FreedomIntelligence/KAUST's AceGPT v2 32B, a base Arabic-focused LLM built on Qwen1.5-32B, part of the AceGPT family spanning 7B to 70B with base and chat variants. Provider: FreedomIntelligence. Released 2024-06-17. Access: Open.
Aggregate Rating 1507 ± 34, rank #756 of 1640 rated models, from 139 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Arabic LLM - Aratrust Illegal | 94.34 | Accuracy (%) | 91 |
| Open Arabic LLM - Arabic MMLU Arabic Language (Grammar) | 53.97 | Accuracy (%) | 85.2 |
| Open Arabic LLM - Arabic MMLU History (High School) | 49.61 | Accuracy (%) | 84.6 |
| Open Arabic LLM - Arabic MMLU Computer Science (Primary School) | 76.84 | Accuracy (%) | 84 |
| Open Arabic LLM - Arabic MMLU HT Public Relations | 62.73 | Accuracy (%) | 82.7 |
| Open Arabic LLM - Arabic MMLU Accounting (University) | 66.22 | Accuracy (%) | 81.5 |
| Open Arabic LLM - Arabic MMLU Driving Test | 72.5 | Accuracy (%) | 80.6 |
| QIMMA - GAT | 51.4 | Accuracy (%, log-likelihood multiple choice) | 80.5 |
| Open Arabic LLM - Arabic MMLU Civics (Middle School) | 56.36 | Accuracy (%) | 79.3 |
| Open Arabic LLM - Arabic MMLU Computer Science (University) | 75 | Accuracy (%) | 79 |
| Open Arabic LLM - Arabic MMLU Social Science (Primary School) | 77.45 | Accuracy (%) | 79 |
| Open Arabic LLM - Arabic MMLU HT Global Facts | 42 | Accuracy (%) | 78.1 |
Interactive version: theaggregate.ai/model?slug=acegpt-v2-32b · How It Works · Data refreshed daily, snapshot 2026-10-13.