ScratchWorld - Debug — leaderboard
Metric: Success Rate (%). Source: huggingface.co. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 2.5 Pro | 75 |
| 2 | Claude Sonnet 4.5 | 75 |
| 3 | DeepSeek V3.2 | 75 |
| 4 | GPT-5 | 65 |
| 5 | Qwen 3 Coder 30B A3B Instruct | 30 |
| 6 | Qwen 3 VL 32B Instruct | 10 |
| 7 | Qwen 2.5 Coder 7B Instruct | 5 |
| 8 | uitars-1.5-7B | 0 |
Interactive version: aibenchmarks.dev/benchmark?slug=scratchworld-debug · How the rankings work · Data refreshed daily, snapshot 2026-07-17.