TableVista - Long Table: leaderboard

Accuracy on the long tables (Web-style rendering) of TableVista (2026), a multimodal table reasoning benchmark in which models answer questions about rendered table images directly with thinking disabled, scored by exact match backed by a semantic-equivalence check. Percent; higher is better.

Metric: Accuracy (%) on the 700 long tables of the 3,000 TableVista questions, Web rendering, direct-output prompt with thinking disabled (reasoning effort none for the GPT models), exact match with a GPT-5-mini semantic-equivalence check on exact-match failures; higher is better. Source: arxiv.org. Saturation forecast: Around February 2027. 29 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-5.4Non-reasoning68.9#135
2Gemma 4 31B (IT)54.4#350
3Llama 4 Maverick Instruct FP852.3#1023
4Qwen 2.5 VL 72B Instruct52.1#841
5Qwen 3.5 27BNon-reasoning50.9#375
6Llama 4 Scout Instruct49.7#1579
7Qwen 3.5 122B A10BNon-reasoning49.4#346
8GPT-5.4 MiniNon-reasoning48.7#406
9Qwen 2.5 VL 32B Instruct48.3#1062
10InternVL3.5-14B46.4#1504
11Gemma 4 26B A4B (IT)44.1#556
12Gemma 3 27B (IT)43.7#1324
13InternVL3.5-30B-A3B43.3#1267
14Qwen 3 VL 30B A3B Instruct42.4#895
15Qwen 3 VL 8B Instruct41.9#1029

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

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