TableVision - Counting: leaderboard

TableVision (2026) category of questions over images of hierarchical tables: L2 reasoning, counting cells (92 questions). Exact-match accuracy (%), zero-shot; higher is better.

Metric: Exact-match accuracy (%) on L2 reasoning, counting cells (92 questions), zero-shot, the model answers directly from the table image; TableVision test set; higher is better. Source: arxiv.org. Saturation forecast: Around July 2027. 10 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Glyph67.4
2Ovis2.5-9B58.7#1225
3Qwen 3 VL 8B Instruct41.3#1029
4GLM-4.1V-9B40.2#1125
5HiPPO39.1
6MiniCPM-V-2.637#3267
7DeepSeek-VL234.8#2927
8Phi-3.5-Vision Instruct31.5#6268
9InternVL3-8B26.1#1782
10LLaVA-OneVision-7B25#2981

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

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