VisEditBench (TikZ diagram editing)
Models edit TikZ code to follow instructions drawn on a diagram. A judge scores adherence, preservation of other content, and readability.
The authors list 29 code editing model entries for 694 diagram edits. The full-score rate counts outputs that earn five on all three image criteria.
- Made by
- Tohoku University and collaborators
- Released
- September 2026
- Kind
- Spatial reasoning and counting
- Level
- University
- Scored
- Graded by a model
- Answers
- Code
- Tasks
- All public
- Measures
- Capability
- Language
- English
- Interaction
- Single turn
- Refresh
- One fixed set
- Run by
- Anyone
- In views
- Vision
- Strongest model tested
- GPT-5.4 #28 on the capability index
- Last model added
- April 2026
- Scores refreshed
- November 20th, 2026
Gemini 3.1 Pro leads with 83.6% among models released by April 2026
Best score per model from the benchmark's own leaderboard, read on November 20th, 2026. Where a model ran with several settings or harnesses, we keep its best. Every run is in the next chart.
The newest model with a score here came out in April 2026, and nobody has tested a newer one. Among the 20 highest-ranked models on the capability index, these have no score yet:
- GPT-6.2 Astra
- Claude Fable 5.2
-
Gemini 4 Carbon
- GPT-6 Astra
- and 16 more
- 83.6
Gemini 3.1 Pro
- 71.8
Gemini 3 Flash
- 66.3
Qwen3.5 397B-A17B
- 59.9 Claude Opus 4.6
- 53.9
Qwen3.5-35B-A3B
- 47.3 GPT-5.4
- 44.8
Gemini 3.1 Flash-Lite
- 43.5
Gemma 4 31B IT
- 42.4 Claude Sonnet 4.5
- 42.1
Qwen3-VL-30B-A3B-Thinking
- 40.5 GLM-4.6V
- 38.3 GLM-4.5V
- 38.0
Qwen3.5-9B
- 36.0
Qwen3-VL-235B-A22B-Instruct
- 35.7
Qwen3-VL-235B-A22B
- 30.8
Qwen3-VL-8B-Thinking
- 30.0
Qwen3-VL-4B-Thinking
- 28.1
Qwen3.5-4B
- 25.2 Claude Haiku 4.5
- 18.7
Qwen3-VL-30B-A3B-Instruct
- 18.2 GPT-5.4 Mini
- 17.9
Qwen3-VL-8B-Instruct
- 12.8
Qwen3-VL-4B-Instruct
- 6.05
Gemma 4 E4B
- 4.61 GPT-5.4 Nano
- 1
Gemini 3.1 Pro 83.6%
- 2
Gemini 3 Flash 71.8%
- 3
Qwen3.5 397B-A17B Thinking 66.3%
- 4 Claude Opus 4.6 59.9%
- 5
Qwen3.5-35B-A3B Thinking 53.9%
- 6 GPT-5.4 47.3%
- 7
Gemini 3.1 Flash-Lite 44.8%
- 8
Gemma 4 31B IT 43.5%
- 9 Claude Sonnet 4.5 42.4%
- 10
Qwen3-VL-30B-A3B-Thinking 42.1%
- 11 GLM-4.6V 40.5%
- 12 GLM-4.5V 38.3%
- 13
Qwen3.5-9B Thinking 38%
- 14
Qwen3-VL-235B-A22B-Instruct 36%
- 15
Qwen3-VL-235B-A22B thinking 35.7%
- 16
Qwen3-VL-8B-Thinking 30.8%
- 17
Qwen3-VL-4B-Thinking 30%
- 18
Qwen3.5-4B Thinking 28.1%
- 19 Claude Haiku 4.5 25.2%
- 20
Qwen3-VL-30B-A3B-Instruct 18.7%
- 21 GPT-5.4 Mini 18.2%
- 22
Qwen3-VL-8B-Instruct 17.9%
- 23
Qwen3-VL-4B-Instruct 12.8%
- 24
Gemma 4 E4B 6.05%
- 25 GPT-5.4 Nano 4.61%
Source: Official VisEditBench (TikZ diagram editing) leaderboard
| # | Model | Score | Setting | Date | Source |
|---|---|---|---|---|---|
| 1 | | 83.6% | Feb 2026 | Official leaderboard | |
| 2 | | 71.8% | Dec 2025 | Official leaderboard | |
| 3 | | 66.3% | Thinking | Feb 2026 | Official leaderboard |
| 4 | Claude Opus 4.6 | 59.9% | Feb 2026 | Official leaderboard | |
| 5 | | 53.9% | Thinking | Feb 2026 | Official leaderboard |
| 6 | GPT-5.4 | 47.3% | Mar 2026 | Official leaderboard | |
| 7 | | 44.8% | Mar 2026 | Official leaderboard | |
| 8 | | 43.5% | Apr 2026 | Official leaderboard | |
| 9 | Claude Sonnet 4.5 | 42.4% | Sep 2025 | Official leaderboard | |
| 10 | | 42.1% | Oct 2025 | Official leaderboard | |
| 11 | GLM-4.6V | 40.5% | Dec 2025 | Official leaderboard | |
| 12 | GLM-4.5V | 38.3% | Aug 2025 | Official leaderboard | |
| 13 | | 38% | Thinking | Feb 2026 | Official leaderboard |
| 14 | | 36% | Sep 2025 | Official leaderboard | |
| 15 | | 35.7% | thinking | Sep 2025 | Official leaderboard |
| 16 | | 30.8% | Oct 2025 | Official leaderboard | |
| 17 | | 30% | Oct 2025 | Official leaderboard | |
| 18 | | 28.1% | Thinking | Feb 2026 | Official leaderboard |
| 19 | Claude Haiku 4.5 | 25.2% | Oct 2025 | Official leaderboard | |
| 20 | | 18.7% | Oct 2025 | Official leaderboard | |
| 21 | GPT-5.4 Mini | 18.2% | Mar 2026 | Official leaderboard | |
| 22 | | 17.9% | Oct 2025 | Official leaderboard | |
| 23 | | 12.8% | Oct 2025 | Official leaderboard | |
| 24 | | 6.05% | Mar 2026 | Official leaderboard | |
| 25 | GPT-5.4 Nano | 4.61% | Mar 2026 | Official leaderboard |
Scores at each effort level
Models the leaderboard ran at two or more effort levels. We don’t have enough cost data here to say what the extra effort costs.
Source: Official VisEditBench (TikZ diagram editing) leaderboard
| Model | Effort | Harness | Score | Cost per task | Output tokens |
|---|---|---|---|---|---|
| 83.6% | |||||
| 71.8% | |||||
| Thinking | 66.3% | ||||
| Claude Opus 4.6 | 59.9% | ||||
| Thinking | 53.9% | ||||
| GPT-5.4 | 47.3% | ||||
| 44.8% | |||||
| 43.5% | |||||
| Claude Sonnet 4.5 | 42.4% | ||||
| 42.1% | |||||
| GLM-4.6V | 40.5% | ||||
| GLM-4.5V | 38.3% | ||||
| 38% | |||||
| Thinking | 38% | ||||
| 36% | |||||
| thinking | 35.7% | ||||
| 31.1% | |||||
| 30.8% | |||||
| 30% | |||||
| Thinking | 28.1% | ||||
| Claude Haiku 4.5 | 25.2% | ||||
| 18.7% | |||||
| GPT-5.4 Mini | 18.2% | ||||
| 17.9% | |||||
| 14.1% | |||||
| 12.8% | |||||
| 12.3% | |||||
| 6.05% | |||||
| GPT-5.4 Nano | 4.61% |
Number one over time
Lab claims count from the day the card came out, independent runs from the model’s release.
Source: Official VisEditBench (TikZ diagram editing) leaderboard
Scores over time
All the scores we have, one dot each, so a model tested at several settings appears more than once.
Source: Official VisEditBench (TikZ diagram editing) leaderboard
No lab model card reports it yet
We read the headline results of 540 launch posts from 57 labs, and none of them list this benchmark.
More in Vision
- MMMU In use 11.5K college-level multimodal questions across 30 subjects. Multiple-choice and open accuracy.
Qwen 3.6 Plus 86%
- MMMU-Pro In use Harder MMMU: college-level multimodal questions with more options and vision-only inputs to remove text shortcuts. GPT-5.6 Sol 84.6%
- CharXiv In use Chart understanding on real arXiv figures. Reasoning questions require synthesizing information from complex charts. Claude Fable 5 93.5%
- MathVista In use Math reasoning in visual contexts (charts, geometry, puzzles) from 31 datasets. Answer accuracy.
Seed2.1 Pro 90.7%
- MATH-Vision (MathVision) In use 3,040 competition math problems with visual context. Answer accuracy.
Qwen 3.8 Max 97.7%
- MMStar In use Models answer 1,500 human-curated image questions that cannot be solved without the image, across 6 core capabilities. Scored by accuracy.
Qwen 3.8 Max 85.9%
Data audit
9 of 10 checks pass, 1 warning. Look first at the Newest score check. The newest came 7 months ago, in April 2026, for Gemma 4 31B IT. Every benchmark
Catalog record 2 pass
- Pass Task access All tasks are public.
- Pass Confidence Checked against a primary source.
Scores 1 warning 3 pass 1 note
- Pass Independent runs Scores from the official leaderboard (25 entries).
- Pass Models scored 25 models we track.
- Warning Newest score The newest came 7 months ago, in April 2026, for Gemma 4 31B IT.
- Note Frontier models None of the top 10 has a score. The strongest tested is GPT-5.4 (#28).
- Pass Headroom The best is 83.6%, under 85%, so it still separates models.
Leaderboard import 3 pass
- Pass Importer We read the board on viseditbench.github.io: 29 rows for models we track at the last read.
- Pass Last read November 20th, 2026.
- Pass Read worked The last read worked.
Consistency 1 pass 1 note
- Note Lab claims No model has both a lab number and an independent run.
- Pass Known issues No published audit found problems.