granite-4.0-h-3B-ar: benchmark results
Provider: IBM. Access: Open.
Aggregate Rating 1422 ± 47, rank #1133 of 1651 rated models, from 50 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| QIMMA - ArabLegalQA | 74.63 | BERTScore F1 (%) | 100 |
| QIMMA - ArabCulture | 69.97 | Accuracy (%, log-likelihood multiple choice) | 85.4 |
| QIMMA - MedAraBench | 39.82 | Accuracy (%, log-likelihood multiple choice) | 65.9 |
| QIMMA - MedArabiQ QA | 52.66 | BERTScore F1 (%) | 64.8 |
| QIMMA - 3LM STEM | 80.46 | Accuracy (%, log-likelihood multiple choice) | 64.6 |
| QIMMA - AraTrust | 83.14 | Accuracy (%, log-likelihood multiple choice) | 64.6 |
| QIMMA - MedArabiQ MCQ | 38.16 | Accuracy (%, log-likelihood multiple choice) | 61 |
| QIMMA - Overall | 56.55 | Mean score across 14 benchmarks (%) | 59.8 |
| QIMMA - GAT | 43.09 | Accuracy (%, log-likelihood multiple choice) | 57.3 |
| QIMMA - MizanQA | 54.53 | Normalized multiple-choice probability score (%) | 53.7 |
| QIMMA - PALMX | 67.38 | Accuracy (%, log-likelihood multiple choice) | 52.4 |
| QIMMA - ArabicMMLU | 60.39 | Accuracy (%, log-likelihood multiple choice) | 51.2 |
Interactive version: theaggregate.ai/model?slug=granite-4-0-h-3b-ar · How It Works · Data refreshed daily, snapshot 2026-10-19.