VectorGym - Sketch2SVG: leaderboard
Metric: Sketch2SVG task score (0-100): SVG code from a hand-drawn sketch, mean of the VLM-judge score, DINO similarity, 100 minus MSE and 100 minus LPIPS of the rendered output against the human-made target (all 0-100), on VectorGym's human-annotated test split (real-world SVGs from SVG-Stack); Gemini 3 Pro judges; higher is better. Source: arxiv.org. Saturation forecast: Around January 2027. 10 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3 Pro | 78.56 | #77 |
| 2 | GPT-5.1 | 75.69 | #131 |
| 3 | Claude Sonnet 4.5 | 73.85 | #138 |
| 4 | GPT-4o | 69.55 | #333 |
| 5 | Qwen 3 VL 235B A22B Instruct | 67.52 | #264 |
| 6 | Gemini 2.5 Flash | 65.45 | #237 |
| 7 | Qwen 3 VL 8B Instruct | 64.82 | #401 |
| 8 | GLM-4.5V | 64.23 | #339 |
| 9 | Qwen 2.5 VL 32B Instruct | 57.91 | #443 |
| 10 | Qwen 2.5 VL 72B Instruct | 55.57 | #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-sketch2svg · How It Works · Data refreshed daily, snapshot 2026-10-11.