WildTableBench - Color-based Reasoning: leaderboard
A WildTableBench (2026) subtype on counting or computing over the cells selected by a color filter plus another condition, asked about real-world table images from the web. Accuracy in percent, free-form answers judged against the reference by GPT-5.2; higher is better.
Metric: Accuracy (%) on the Color-based Reasoning subtype (C5-R), 928 human-verified questions over 402 real-world table images, free-form answers judged by GPT-5.2 against the reference answer, high reasoning effort where configurable; higher is better. Source: arxiv.org. Saturation forecast: Around June 2027. 21 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Setting | Score | Overall rank |
|---|---|---|---|---|
| 1 | Gemini 3 Pro | 51.6 | #145 | |
| 2 | Kimi K2.5 | Thinking | 35.5 | #266 |
| 3 | Gemini 3 Flash | 33.9 | #175 | |
| 4 | Qwen 3 VL 32B (Thinking) | 22.6 | #573 | |
| 5 | GPT-5.2 | High | 22.6 | #185 |
| 6 | Seed 2.0 Pro | High | 22.6 | #222 |
| 7 | Qwen 3 VL 235B A22B (Thinking) | 21 | #459 | |
| 8 | GLM-4.6V | 19.4 | #710 | |
| 9 | O3 | High | 14.5 | #237 |
| 10 | Qwen 3 VL 235B A22B Instruct | 12.9 | #532 | |
| 11 | GPT-5 Mini | High | 12.9 | #364 |
| 12 | Claude Sonnet 4.6 | High | 12.9 | #173 |
| 13 | Claude Opus 4.6 | High | 11.3 | #113 |
| 14 | Qwen 3 VL 32B Instruct | 9.7 | #606 | |
| 15 | Qwen 3 VL 8B (Thinking) | 6.5 | #910 |
No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.
Interactive version: theaggregate.ai/benchmark?slug=wildtablebench-color-based-reasoning · How It Works · Data refreshed daily, snapshot 2026-11-21.