WildTableBench - Hypothetical Condition: leaderboard

A WildTableBench (2026) subtype on deriving a number or truth value under a what-if scenario, 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 Hypothetical Condition subtype (C4-H), 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 February 2027. 21 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3 Pro82.1#145
2GPT-5.2High60.7#185
3Kimi K2.5Thinking53.6#266
4Claude Sonnet 4.6High53.6#173
5Gemini 3 Flash50#175
6Qwen 3 VL 235B A22B (Thinking)50#459
7Claude Opus 4.6High50#113
8Seed 2.0 ProHigh42.9#222
9GPT-5 MiniHigh32.1#364
10Qwen 3 VL 8B (Thinking)25#910
11Qwen 3 VL 32B Instruct21.4#606
12GLM-4.6V17.9#710
13Qwen 3 VL 235B A22B Instruct14.3#532
14Qwen 3 VL 32B (Thinking)10.7#573
15O3High10.7#237

No result here: #1 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=wildtablebench-hypothetical-condition · How It Works · Data refreshed daily, snapshot 2026-11-21.