TableVista - Easy Questions: leaderboard

Accuracy on the easy questions, which need the fewest reasoning steps (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 easy questions (fewest reasoning steps) 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: Estimated already saturated. 29 models tracked. Updated 7 Oct 2026.

Top models

#ModelSettingScoreOverall rank
1GPT-5.4Non-reasoning93.6#138
2Qwen 2.5 VL 72B Instruct90.2#849
3Gemma 4 31B (IT)88.2#354
4Qwen 2.5 VL 32B Instruct84.5#1065
5Llama 4 Maverick Instruct FP884.4#1034
6Qwen 3.5 27BNon-reasoning84.2#380
7InternVL3.5-14B83#1519
8Qwen 3.5 122B A10BNon-reasoning82.6#350
9Llama 4 Scout Instruct81.8#1580
10InternVL3.5-30B-A3B81.6#1269
11Gemma 4 26B A4B (IT)79.7#563
12Qwen 2.5 VL 7B Instruct79#1919
13Qwen 3 VL 30B A3B Instruct78.9#903
14Qwen 3 VL 8B Instruct76.7#1043
15Gemma 3 27B (IT)76.5#1341

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

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