Edit2TikZ - Compilation Success Rate: leaderboard
Edit2TikZ (arXiv 2608.13441v1, Table 3): 1,548 scientific-figure edits where a multimodal model gets a rendered figure and an instruction and must write compilable TikZ for the revised figure. Share of outputs that compile and render in a shared TeX Live environment, percent. Higher is better.
Metric: Compiled and rendered outputs (%). Source: arxiv.org. Saturation forecast: Around February 2027. 14 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 88.37 | #64 |
| 2 | Claude Opus 4.8 | 78.88 | #39 |
| 3 | GPT-5.6 Terra | 77.33 | #40 |
| 4 | GPT-5.6 Sol | 75.65 | #22 |
| 5 | Claude Sonnet 4.6 | 74.1 | #94 |
| 6 | GPT-5.6 Luna | 74.1 | #79 |
| 7 | Qwen 3.6 27B | 69.77 | #182 |
| 8 | Qwen 3.7 Plus | 67.05 | #98 |
| 9 | Seed 2.1 Pro | 65.76 | #73 |
| 10 | gemma-4-E4B-it | 52.39 | #577 |
| 11 | Qwen 3.5 9B | 52.26 | #391 |
| 12 | Qwen 3.5 4B | 45.35 | #514 |
| 13 | Qwen 3.5 2B | 13.24 | #1114 |
| 14 | Gemma 4 E2B (IT) | 6.72 | #779 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=edit2tikz-compilation-success-rate · How It Works · Data refreshed daily, snapshot 2026-11-05.