AceGPT-v2-32B-Chat: benchmark results
FreedomIntelligence/KAUST's AceGPT v2 32B Chat, a dialogue-tuned Arabic-focused LLM fine-tuned from AceGPT-v2-32B, part of the AceGPT family spanning 7B to 70B. Provider: FreedomIntelligence. Released 2024-06-17. Access: Open.
Aggregate Rating 1581 ± 20, rank #511 of 1661 rated models, from 277 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| QIMMA - ArabCulture | 79.79 | Accuracy (%, log-likelihood multiple choice) | 98.8 |
| HELM Arabic - Alghafa - Multiple Choice Facts Truefalse Balanced Task | 97.33 | Exact Match (%) | 97.5 |
| Open Arabic LLM - Madinah QA Arabic Language (Grammar) | 69.59 | Accuracy (%) | 96.3 |
| Open Arabic LLM - Arabic MMLU Arabic Language (Grammar) | 68.22 | Accuracy (%) | 95.7 |
| Open Arabic LLM - Arabic MMLU Civics (High School) | 55.17 | Accuracy (%) | 93.8 |
| Open Arabic LLM - Arabic MMLU Arabic Language (Middle School) | 81.48 | Accuracy (%) | 93.2 |
| Open Arabic LLM - Arabic MMLU History (High School) | 53.55 | Accuracy (%) | 93.2 |
| QIMMA - GAT | 56.04 | Accuracy (%, log-likelihood multiple choice) | 92.7 |
| Open Arabic LLM - Arabic MMLU Civics (Middle School) | 61.86 | Accuracy (%) | 92.6 |
| Open Arabic LLM - Madinah QA Average | 73.77 | Accuracy (%) | 92.6 |
| Open Arabic LLM - Alghafa Multiple Choice Facts Truefalse Balanced Task | 94.67 | Accuracy (%) | 92.3 |
| Open Arabic LLM - Arabic MMLU General Knowledge (Primary School) | 77.16 | Accuracy (%) | 92 |
Interactive version: theaggregate.ai/model?slug=acegpt-v2-32b-chat · How It Works · Data refreshed daily, snapshot 2026-10-23.