Atlas-Chat-9B: benchmark results
Provider: Other. Access: Open.
Aggregate Rating 1443 ± 24, rank #1033 of 1657 rated models, from 30 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| QIMMA - MizanQA | 59.69 | Normalized multiple-choice probability score (%) | 81.7 |
| QIMMA - MedAraBench | 42.52 | Accuracy (%, log-likelihood multiple choice) | 76.8 |
| QIMMA - 3LM STEM | 82.42 | Accuracy (%, log-likelihood multiple choice) | 70.7 |
| QIMMA - MedArabiQ MCQ | 41.21 | Accuracy (%, log-likelihood multiple choice) | 70.7 |
| QIMMA - PALMX | 71.66 | Accuracy (%, log-likelihood multiple choice) | 69.5 |
| QIMMA - AraTrust | 83.33 | Accuracy (%, log-likelihood multiple choice) | 65.9 |
| QIMMA - ArabCulture | 53.27 | Accuracy (%, log-likelihood multiple choice) | 65.9 |
| QIMMA - GAT | 45.35 | Accuracy (%, log-likelihood multiple choice) | 63.4 |
| QIMMA - Overall | 57.27 | Mean score across 14 benchmarks (%) | 63.4 |
| DAR-MGSM - Moroccan - Localized | 60 | Accuracy (%) | 60 |
| DAR-MGSM - Moroccan - Translated | 58.4 | Accuracy (%) | 60 |
| QIMMA - ArabicMMLU | 62.07 | Accuracy (%, log-likelihood multiple choice) | 54.9 |
Interactive version: theaggregate.ai/model?slug=atlas-chat-9b · How It Works · Data refreshed daily, snapshot 2026-10-22.