WildTableBench - Ranking: leaderboard

A WildTableBench (2026) subtype on finding the maximum, minimum or the entity at a given rank, 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 Ranking subtype (C2-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 February 2027. 21 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1Gemini 3 Pro80#145
2Kimi K2.5Thinking62.5#266
3GPT-5.2High45#185
4Gemini 3 Flash42.5#175
5Claude Opus 4.6High42.5#113
6Seed 2.0 ProHigh37.5#222
7Qwen 3 VL 32B (Thinking)25#573
8Claude Sonnet 4.6High25#173
9Qwen 3 VL 235B A22B (Thinking)22.5#459
10Qwen 3 VL 32B Instruct20#606
11Qwen 3 VL 235B A22B Instruct20#532
12GPT-5 MiniHigh20#364
13Qwen 3 VL 8B Instruct10#1029
14Qwen 3 VL 8B (Thinking)10#910
15GLM-4.6V10#710

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

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