Jais 3 Model Cards - Terminal-Bench 2.1 (Avg@3, Jais Runs): leaderboard
Terminal-Bench 2.1 command-line tasks, average of three runs; only the Jais models are included because several reference cells on the cards repeat Artificial Analysis or vendor-reported values. MBZUAI's Institute of Foundation Models reported this on the Jais 3 model cards (released October 15, 2026); this board keeps the 5 Jais 3 runs they report. The cards state that all reported results use high reasoning effort. Vendor-reported. Metric: Accuracy (%). Higher is better. Source: https://huggingface.co/collections/IFM/jais-3-6abba9df4f1173ec7ce2929d
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Rough model projection: around 2026. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Jais-3-375B-A23B | 73.8 | |
| 2 | Jais-3-MoVA-36B-A4B | 67.3 | |
| 3 | Jais-3-32B | 63.7 | |
| 4 | Jais-3-7B | 46.4 | |
| 5 | Jais-3-3.7B | 41.2 |
Interactive version: theaggregate.ai/benchmark?slug=jais-3-model-cards-terminal-bench-2-1-avg-3-jais-runs · How It Works · Data refreshed daily, snapshot 2026-10-19.