EdiTikZ: leaderboard
DaEdiTikZ-Bench scientific figure editing aggregate score on a 0–1 scale, Table 3. The score combines normalized tree-edit, diagram similarity, edit application, source preservation and visual quality; specialist EdiTikZ SFT and RL checkpoints remain distinct.
Metric: Overall Edit Score (0-1). Source: arxiv.org. Saturation forecast: Around January 2027. 21 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 0.69 | #58 |
| 2 | GPT-5.5 | 0.69 | #31 |
| 3 | GPT-5.6 Sol | 0.67 | #19 |
| 4 | GPT-5.4 | 0.62 | #79 |
| 5 | Gemini 3.5 Flash | 0.6 | #60 |
| 6 | Gemini 3.6 Flash | 0.59 | #51 |
| 7 | Qwen 3.5 27B | 0.51 | #187 |
| 8 | Qwen 3.6 27B | 0.51 | #170 |
| 9 | Qwen 3.5 9B | 0.35 | #379 |
| 10 | Qwen 3.5 4B | 0.25 | #475 |
| 11 | Qwen 2.5 VL 7B | 0.24 | #643 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=editikz · How It Works · Data refreshed daily, snapshot 2026-10-20.