TurtleAI: leaderboard
Metric: Success rate (%) on all 823 TurtleAI-DS tasks (102 real, 102 hand-drawn, 619 synthetic), symbolic comparison: the predicted Python turtle program is run in a canonicalizing emulator and succeeds when its normalized drawing matches the reference pixel by pixel (0.95 or 0.92 overlap threshold), greedy decoding, one fixed prompt with the task image; higher is better. Source: arxiv.org. Saturation forecast: Around 2028. 18 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | O4 Mini (Medium) | 15.92 |
| 2 | O3 (Medium) | 13.73 |
| 3 | GPT-4o (2024-11-20) | 9.23 |
| 4 | GPT-5 (Medium) | 9.11 |
| 5 | Qwen 2.5 VL 72B Instruct | 6.8 |
| 6 | Qwen 3 VL 30B A3B Instruct | 6.2 |
| 7 | GPT-4 Turbo | 5.47 |
| 8 | InternVL3-78B | 4.98 |
| 9 | Pixtral-12B | 2.31 |
| 10 | Qwen 2 VL 7B Instruct | 0.12 |
Interactive version: theaggregate.ai/benchmark?slug=turtleai · How It Works · Data refreshed daily, snapshot 2026-09-29.