WildTableBench - Basic Numerical: leaderboard

A WildTableBench (2026) subtype on a single operation such as a sum, difference, average or median over a row or column, 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 Basic Numerical subtype (C2-B), 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 Pro77.1#146
2Claude Opus 4.6High54.3#114
3Kimi K2.5Thinking51.4#272
4GPT-5.2High45.7#190
5Seed 2.0 ProHigh40#226
6Gemini 3 Flash37.1#180
7Qwen 3 VL 235B A22B (Thinking)37.1#462
8Qwen 3 VL 32B (Thinking)37.1#577
9Claude Sonnet 4.6High34.3#179
10Qwen 3 VL 235B A22B Instruct28.6#541
11Qwen 3 VL 8B (Thinking)28.6#924
12GPT-5 MiniHigh28.6#373
13GLM-4.6V25.7#721
14Qwen 3 VL 32B Instruct17.1#607
15O3High14.3#242

No result here: #2 Claude Fable 5.2, #3 Gemini 4 Neon, #4 GPT-6.2 Astra.

Interactive version: theaggregate.ai/benchmark?slug=wildtablebench-basic-numerical · How It Works · Data refreshed daily, snapshot 2026-11-22.