VectorGym - SVG Captioning: leaderboard
Metric: SVG Captioning task score (0-100): description of an SVG from its code, mean of the VLM-judge score, BGE-M3 similarity and ROUGE-L against the human caption (all 0-100), on VectorGym's human-annotated test split (real-world SVGs from SVG-Stack); higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 55.5 | #77 |
| 2 | GPT-5.1 | 51.38 | #131 |
| 3 | Claude Sonnet 4.5 | 50.15 | #138 |
| 4 | Gemini 2.5 Flash | 45.83 | #237 |
| 5 | GPT-4o | 44.72 | #333 |
| 6 | Qwen 3 VL 235B A22B Instruct | 41.96 | #264 |
| 7 | Qwen 2.5 VL 32B Instruct | 40.28 | #443 |
| 8 | GLM-4.5V | 39.24 | #339 |
| 9 | Qwen 3 VL 8B Instruct | 36.78 | #401 |
| 10 | Qwen 2.5 VL 72B Instruct | 23.13 | #364 |
No result here: #3 Claude Opus 5.5, #5 GPT-6 Astra, #8 Claude Fable 5.1.
Interactive version: theaggregate.ai/benchmark?slug=vectorgym-svg-captioning · How It Works · Data refreshed daily, snapshot 2026-10-11.