ALLaM-7B-Instruct-preview: benchmark results
SDAIA's ALLaM-7B-Instruct-preview, an Arabic-English instruction-tuned model trained from scratch on 4T English tokens followed by 1.2T mixed Arabic/English tokens. Provider: SDAIA. Released 2025-02-13. Access: Open.
Aggregate Rating 1521 ± 13, rank #708 of 1635 rated models, from 337 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Open Arabic LLM - Aratrust Ethics | 83.33 | Accuracy (%) | 99.1 |
| Open Arabic LLM - Aratrust Trustfulness | 82.05 | Accuracy (%) | 99.1 |
| HELM Arabic - Arabic Mmlu - Philosophy (High School) | 82.05 | Exact Match (%) | 97.5 |
| QIMMA - AraDiCE-Culture | 78.89 | Accuracy (%, log-likelihood multiple choice) | 97 |
| Open Arabic LLM - Arabic MMLU Computer Science (Middle School) | 96.3 | Accuracy (%) | 96.3 |
| Open Arabic LLM - Arabic MMLU Arabic Language (Primary School) | 81.75 | Accuracy (%) | 95.4 |
| Open Arabic LLM - Arabic MMLU Islamic Studies (Primary School) | 84.78 | Accuracy (%) | 93.8 |
| Open Arabic LLM - Arabic MMLU History (Middle School) | 73.4 | Accuracy (%) | 93.5 |
| Open Arabic LLM - Aratrust Offensive | 94.2 | Accuracy (%) | 93.5 |
| Open Arabic LLM - Alghafa Multiple Choice Facts Truefalse Balanced Task | 94.67 | Accuracy (%) | 92.3 |
| Open Arabic LLM - Arabic MMLU Islamic Studies (Middle School) | 75.21 | Accuracy (%) | 92.3 |
| QIMMA - ArabicMMLU | 72.54 | Accuracy (%, log-likelihood multiple choice) | 90.2 |
Interactive version: theaggregate.ai/model?slug=allam-7b-instruct-preview · How It Works · Data refreshed daily, snapshot 2026-10-15.