Jais 3 Model Cards - WildClawBench (Mean Score@3): leaderboard
WildClawBench in-the-wild agentic tasks, mean score over three runs, as labelled on the Jais-3-3.7B card (which does not name the task subset). MBZUAI's Institute of Foundation Models reported this on the Jais-3-3.7B model card (released October 15, 2026), comparing it with size-matched reference models. The card states that all reported results use high reasoning effort. Vendor-reported. Metric: Mean Score@3. Higher is better. Source: https://huggingface.co/collections/IFM/jais-3-6abba9df4f1173ec7ce2929d
Metric: Mean Score@3. Source: huggingface.co. Saturation forecast: Around January 2027. 5 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen 3.5 4B | 33.3 | #473 |
| 2 | Gemma 4 E2B | 18.2 | #1076 |
| 3 | granite-4.0-h-3B-ar | 8.2 | #1133 |
| 4 | Falcon-H1-3B-Instruct | 0.1 | #1425 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=jais-3-model-cards-wildclawbench-mean-score-3 · How It Works · Data refreshed daily, snapshot 2026-10-19.