ScratchWorld - Overall: leaderboard
ScratchWorld (Zhang et al., 2026) gives multimodal agents 83 Scratch programming tasks in four categories (Create, Debug, Extend, Compute), each checked by unit tests run in the Scratch VM. These are composite-mode results: the agent reads the program as pseudocode and edits it through semantic commands such as adding or connecting blocks, which separates program logic from GUI control. The score is the percentage of all 83 tasks on which every unit test passes; higher is better.
Metric: Success Rate (%). Source: huggingface.co. Saturation forecast: Around January 2027. 9 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Claude Sonnet 4.5 | 78.31 | #142 |
| 2 | Gemini 2.5 Pro | 72.29 | #152 |
| 3 | GPT-5 | 62.65 | #98 |
| 4 | DeepSeek V3.2 | 61.45 | #205 |
| 5 | Qwen 3 Coder 30B A3B Instruct | 9.64 | #415 |
| 6 | Qwen 2.5 Coder 7B Instruct | 1.2 | #951 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=scratchworld-overall · How It Works · Data refreshed daily, snapshot 2026-10-20.