WildTableBench - Color-based Reasoning: leaderboard
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 January 2028. 21 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Gemini 3 Pro | 51.6 |
| 2 | Kimi K2.5 (Thinking) | 35.5 |
| 3 | Gemini 3 Flash | 33.9 |
| 4 | GPT-5.2 (High) | 22.6 |
| 5 | Qwen 3 VL 32B (Thinking) | 22.6 |
| 6 | Seed 2.0 Pro (High) | 22.6 |
| 7 | Qwen 3 VL 235B A22B (Thinking) | 21 |
| 8 | GLM-4.6V | 19.4 |
| 9 | O3 (High) | 14.5 |
| 10 | Qwen 3 VL 235B A22B Instruct | 12.9 |
| 11 | GPT-5 Mini (High) | 12.9 |
| 12 | Claude Sonnet 4.6 (High) | 12.9 |
| 13 | Claude Opus 4.6 (High) | 11.3 |
| 14 | Qwen 3 VL 32B Instruct | 9.7 |
| 15 | Qwen 3 VL 8B (Thinking) | 6.5 |
Interactive version: theaggregate.ai/benchmark?slug=wildtablebench-color-based-reasoning · How It Works · Data refreshed daily, snapshot 2026-10-07.