TableVista - Screen Capture: leaderboard

Accuracy in the vision-only Screen Capture setting, with question, context and table in one image, of all questions in 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 3,000 TableVista questions under the vision-only Screen Capture setting (question, context and table rendered in one image), 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-reasoning69.4#135
2Qwen 2.5 VL 72B Instruct57.4#841
3Qwen 3.5 27BNon-reasoning55.1#375
4Qwen 3.5 122B A10BNon-reasoning54.3#346
5Llama 4 Maverick Instruct FP853.4#1023
6Qwen 2.5 VL 32B Instruct51.1#1062
7Llama 4 Scout Instruct46.7#1579
8Qwen 3.6 35B A3BNon-reasoning46.1#481
9Qwen 3 VL 8B Instruct45.3#1029
10InternVL3.5-14B42.7#1504
11Gemma 4 31B (IT)42.2#350
12GPT-5.4 MiniNon-reasoning42#406
13Qwen 2.5 VL 7B Instruct41.6#1902
14Qwen 3 VL 30B A3B Instruct41.1#895
15Gemma 3 27B (IT)40.1#1324

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

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