Ducky Bench (Africa): leaderboard
ELO-ranked SVG generation benchmark where models recreate a reference Africa image as vector graphics, judged by human voters on visual accuracy.
Metric: ELO. Source: ducky-bench.joinity.site. Saturation forecast: Around March 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | DeepSeek R1 | 1267 | #251 |
| 2 | GPT-5 (High) | 1253 | #98 (GPT-5) |
| 3 | Gemini 3 Pro | 1185 | #80 |
| 4 | Claude Opus 4.1 (Thinking) | 1150 | #140 (Claude Opus 4.1) |
| 5 | Gemini 2.5 Pro | 991 | #152 |
| 6 | O3 (High) | 962 | #124 (O3) |
| 7 | Grok 4 | 874 | #181 |
| 8 | Claude Sonnet 4.5 | 846 | #142 |
| 9 | GPT-4o | 839 | #343 |
| 10 | GPT-5 Mini (High) | 781 | #186 (GPT-5 Mini) |
| 11 | Llama 4 Maverick | 641 | #464 |
| 12 | Kimi K2 | 593 | #243 |
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-africa · How It Works · Data refreshed daily, snapshot 2026-10-20.