AIC-1: benchmark results
The Applied Innovation Center's open-weight AIC-1 (September 2025), a 32B-parameter Arabic-specialized fine-tune of Qwen2.5-32B-Instruct, tuned via instruction fine-tuning and DPO. Provider: Applied Innovation Center. Released 2025-09-10. Access: Open.
Aggregate Rating 1688 ± 33, rank #223 of 1635 rated models, from 139 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Arabic LLM - Alghafa Multiple Choice Grounded Statement Soqal Task | 96.67 | Accuracy (%) | 100 |
| QIMMA - AraTrust | 90.61 | Accuracy (%, log-likelihood multiple choice) | 100 |
| QIMMA - GAT | 62.11 | Accuracy (%, log-likelihood multiple choice) | 100 |
| Open Arabic LLM - Arabic MMLU Arabic Language (Grammar) | 72.6 | Accuracy (%) | 99.4 |
| Open Arabic LLM - Madinah QA Arabic Language (Grammar) | 74.79 | Accuracy (%) | 99.4 |
| Open Arabic LLM - Madinah QA Average | 78 | Accuracy (%) | 99.4 |
| Open Arabic LLM - Aratrust Trustfulness | 82.05 | Accuracy (%) | 99.1 |
| Open Arabic LLM - Alghafa Average | 79 | Accuracy (%) | 98.1 |
| Open Arabic LLM - Alghafa Meta AR MSA | 91.73 | Accuracy (%) | 97.5 |
| Open Arabic LLM - Arabic MMLU HT US Foreign Policy | 87 | Accuracy (%) | 97.5 |
| Open Arabic LLM - Arabic MMLU HT Jurisprudence | 78.7 | Accuracy (%) | 97.2 |
| Open Arabic LLM - Arabic MMLU Geography (Primary School) | 82.46 | Accuracy (%) | 96.9 |
Interactive version: theaggregate.ai/model?slug=aic-1 · How It Works · Data refreshed daily, snapshot 2026-10-15.