WildTableBench - Excel Formula: leaderboard

A WildTableBench (2026) subtype on writing the Excel formula that computes a requested operation over a cell range, 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 Excel Formula subtype (C1-F), 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 Pro61.1#146
2Seed 2.0 ProHigh58.3#226
3Claude Opus 4.6High52.8#114
4GPT-5.2High50#190
5Gemini 3 Flash44.4#180
6Qwen 3 VL 235B A22B (Thinking)38.9#462
7Kimi K2.5Thinking36.1#272
8Claude Sonnet 4.6High33.3#179
9Qwen 3 VL 32B Instruct25#607
10Qwen 3 VL 32B (Thinking)25#577
11O3High22.2#242
12Qwen 3 VL 235B A22B Instruct13.9#541
13GLM-4.6V13.9#721
14GPT-5 MiniHigh13.9#373
15Qwen 3 VL 8B Instruct8.3#1043

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

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