TableVista - Text-Mixed: leaderboard

Accuracy on the tables paired with textual context (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 300 tables paired with textual context 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-reasoning86.7#138
2GPT-5.4 MiniNon-reasoning61#409
3Qwen 2.5 VL 72B Instruct58.3#849
4Llama 4 Maverick Instruct FP855.7#1034
5Gemma 4 31B (IT)54#354
6Qwen 2.5 VL 32B Instruct49.7#1065
7InternVL3.5-30B-A3B49#1269
8Qwen 3 VL 30B A3B Instruct48.7#903
9Qwen 3.5 122B A10BNon-reasoning48.3#350
10InternVL3.5-14B48#1519
11Llama 4 Scout Instruct47.7#1580
12Gemma 4 26B A4B (IT)45.7#563
13Qwen 3.5 27BNon-reasoning45.3#380
14Gemma 3 27B (IT)44#1341
15Qwen 3 VL 8B Instruct44#1043

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

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