Jais-3-32B: benchmark results
Provider: Inception (G42), Cerebras & MBZUAI. Released 2026-10-15. Access: Open.
Aggregate Rating 1720 ± 38, rank #268 of 10916 rated models, from 39 benchmark results.
Strongest benchmark results
| Benchmark | Score | Metric | Percentile |
|---|---|---|---|
| Jais 3 Model Cards - BrowseComp-Plus | 71.9 | Accuracy (%) | 100 |
| Jais 3 Model Cards - QIAS-1 (IFM-audited) | 95.9 | Accuracy (%) | 100 |
| Jais 3 Model Cards - QADI Dialect Identification | 70.6 | Accuracy (%) | 96.4 |
| Jais 3 Model Cards - Arabic IFEval | 80 | Accuracy (%) | 94.6 |
| Jais 3 Model Cards - GAIA v2 lit (Arabic) | 51.5 | Pass@1 (%) | 93.5 |
| Jais 3 Model Cards - AraDiCE-Culture (IFM-audited) | 88.3 | Accuracy (%) | 92.9 |
| Jais 3 Model Cards - AraGen v3 - Conciseness | 28.1 | 3C3H Conciseness (%) | 92.9 |
| Jais 3 Model Cards - Arabic GSM8K (IFM-audited) | 94.3 | Accuracy (%) | 92.9 |
| Jais 3 Model Cards - Arabic Legal (HELM, IFM-audited) | 73.4 | Accuracy (%) | 92.9 |
| Jais 3 Model Cards - FannOrFlop MCQ | 71.8 | Accuracy (%) | 92.9 |
| Jais 3 Model Cards - GDPval (75% Threshold) | 75.5 | Accuracy (%) | 91.3 |
| Jais 3 Model Cards - Alyah (IFM-audited) | 83 | Accuracy (%) | 91.1 |
Interactive version: theaggregate.ai/model?slug=jais-3-32b · How It Works · Data refreshed daily, snapshot 2026-11-08.