Ducky Bench (Stabby Quack): leaderboard
ELO-ranked SVG generation benchmark where models recreate a reference duck image as vector graphics, judged by human voters on visual accuracy.
Metric: ELO. Source: ducky-bench.joinity.site. Saturation forecast: Around January 2027. 29 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-4.1 | 1423 | #248 |
| 2 | Claude Opus 4.1 (Thinking) | 1414 | #140 (Claude Opus 4.1) |
| 3 | Gemini 2.5 Pro | 1374 | #152 |
| 4 | O3 (High) | 1363 | #124 (O3) |
| 5 | Gemini 3 Pro | 1348 | #80 |
| 6 | GPT-5 (High) | 1346 | #98 (GPT-5) |
| 7 | GPT-4.5 | 1284 | #274 |
| 8 | Claude Opus 4 (Thinking) | 1197 | #163 (Claude Opus 4) |
| 9 | Claude Sonnet 4.5 | 1187 | #142 |
| 10 | GPT-5 Mini (High) | 1183 | #186 (GPT-5 Mini) |
| 11 | Gemini 2.5 Flash | 1092 | #245 |
| 12 | GPT-4o | 1072 | #343 |
| 13 | O1 | 1038 | #183 |
| 14 | GPT-5.1 (High) | 988 | #135 (GPT-5.1) |
| 15 | Claude Sonnet 4 (Thinking) | 976 | #199 (Claude Sonnet 4) |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=ducky-bench-stabby-quack · How It Works · Data refreshed daily, snapshot 2026-10-20.