KidGym - Execution: leaderboard
Metric: Capability score (0-100): 100 times the mean, over the KidGym tasks that measure the capability, of each task's success rate weighted 0.2, 0.3 and 0.5 across difficulty levels 1 to 3 (100 zero-shot rounds per task and level in randomized 2D grid scenes; at each step the model picks one action from a shuffled lettered list); execution is measured by the Classification task (placing items into the containers an instruction names); higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 13 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 100 | #145 |
| 2 | GPT-5 | 95 | #91 |
| 3 | O3 | 95 | #121 |
| 4 | Claude 3.7 Sonnet | 88 | #241 |
| 5 | Gemini 2.5 Flash | 63 | #237 |
| 6 | InternVL3-38B | 28 | #395 |
| 7 | Qwen 2.5 VL 32B Instruct | 26 | #443 |
| 8 | GPT-4o | 23 | #333 |
| 9 | Qwen 2.5 VL 72B Instruct | 19 | #364 |
| 10 | InternVL3-78B | 17 | #345 |
| 11 | InternVL3-8B | 7 | #606 |
| 12 | Qwen 2.5 VL 7B Instruct | 7 | #643 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=kidgym-execution · How It Works · Data refreshed daily, snapshot 2026-10-11.