TableVista - Simulated Photo: leaderboard

Accuracy in the vision-only Simulated Photo setting, a screen capture with moire and perspective distortion, 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 Simulated Photo setting (screen capture with moire and perspective distortion), 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-reasoning67.3#135
2Qwen 2.5 VL 72B Instruct54#841
3Qwen 3.5 27BNon-reasoning53.5#375
4Qwen 3.5 122B A10BNon-reasoning53.5#346
5Llama 4 Maverick Instruct FP851.8#1023
6Qwen 2.5 VL 32B Instruct49.8#1062
7Llama 4 Scout Instruct47.2#1579
8Qwen 3 VL 8B Instruct44.6#1029
9Qwen 3.6 35B A3BNon-reasoning44.5#481
10Gemma 4 31B (IT)44.2#350
11Qwen 3 VL 30B A3B Instruct42.4#895
12InternVL3.5-14B41.6#1504
13Qwen 2.5 VL 7B Instruct40#1902
14Qwen 3.5 9BNon-reasoning37.9#828
15GPT-5.4 MiniNon-reasoning37.7#406

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

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