TableVista - Medium Questions: leaderboard

Accuracy on the medium-difficulty questions (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 medium questions 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-reasoning80.1#135
2GPT-5.4 MiniNon-reasoning56.1#406
3Gemma 4 31B (IT)55.6#350
4Qwen 2.5 VL 72B Instruct54.3#841
5Llama 4 Maverick Instruct FP853.7#1023
6Qwen 2.5 VL 32B Instruct49#1062
7Qwen 3.5 27BNon-reasoning47.5#375
8Llama 4 Scout Instruct47#1579
9Qwen 3.5 122B A10BNon-reasoning45.7#346
10InternVL3.5-14B42.1#1504
11Gemma 4 26B A4B (IT)41.4#556
12InternVL3.5-30B-A3B41.3#1267
13Gemma 3 27B (IT)39.5#1324
14Qwen 3 VL 30B A3B Instruct39.4#895
15Qwen 2.5 VL 7B Instruct38.6#1902

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

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