TurtleAI - Synthetic: leaderboard
The synthetic set of TurtleAI (2026): 619 turtle graphics programming tasks generated from the real tasks by code mutation and manually selected for quality. Success means the generated program, run in a canonicalizing emulator, draws a pattern that matches the reference pixel by pixel; percent, higher is better.
Metric: Success rate (%) on the 619 TurtleAI-DSSyn tasks (generated from the real set and manually selected), 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 February 2028. 18 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | O4 Mini (Medium) | 10.5 | #353 (O4 Mini) |
| 2 | O3 (Medium) | 10.18 | #228 (O3) |
| 3 | GPT-5 (Medium) | 7.43 | #167 (GPT-5) |
| 4 | GPT-4o (2024-11-20) | 5.82 | #866 |
| 5 | Qwen 2.5 VL 72B Instruct | 3.55 | #846 |
| 6 | InternVL3-78B | 3.23 | #738 |
| 7 | GPT-4 Turbo | 3.23 | #1001 |
| 8 | Qwen 3 VL 30B A3B Instruct | 2.91 | #871 |
| 9 | Qwen 2 VL 72B Instruct | 1.45 | #1403 |
| 10 | Pixtral-12B | 0.97 | #2476 |
| 11 | Qwen 2 VL 7B Instruct | 0 | #2553 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=turtleai-synthetic · How It Works · Data refreshed daily, snapshot 2026-11-13.