Jais 3 Model Cards - WildClawBench (English Text-Only Subset): leaderboard
WildClawBench in-the-wild agentic tasks, English text-only subset, mean score over three runs. MBZUAI's Institute of Foundation Models reported this on the Jais 3 model cards (released October 15, 2026), comparing each Jais 3 size with size-matched reference models; this board collects the 4 cards that report it (7B, 32B, MoVA-36B-A4B and 375B-A23B). The cards state 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. 17 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Qwen3.8-Flash-Next | 59 | #69 |
| 2 | Qwen 3.8 27B | 52.1 | #104 |
| 3 | Qwen 3.5 397B A17B | 46.6 | #144 |
| 4 | Gemma 4 31B (IT) | 44.1 | #174 |
| 5 | Muse Glimmer 30B | 43.2 | #178 |
| 6 | Qwen 3.5 9B | 34.4 | #377 |
| 7 | Gemma 4 12B (IT) | 29 | #371 |
| 8 | gemma-4-E4B-it | 22.3 | #566 |
| 9 | Falcon-H1-34B-Instruct | 10.7 | #666 |
| 10 | Falcon-H1-7B-Instruct | 9.4 | #1043 |
| 11 | Fanar-1-9B-Instruct | 0 | #633 |
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-english-text-only-subset · How It Works · Data refreshed daily, snapshot 2026-10-19.