TriViewBench - Object Counting: leaderboard

Metric: Accuracy (%) over the 7,692 Object Counting questions (objects visible in each view and in total), all four complexity levels; three orthographic views (front, side, top) of synthetic 3D tower scenes, direct prompting with a concise answer within 128 tokens, greedy decoding; higher is better. Source: arxiv.org. Saturation forecast: Estimated already saturated. 18 models tracked.

Top models

#ModelScore
1Gemini 2.5 Flash78.59
2Claude 3.7 Sonnet68.23
3GPT-4o58.52
4Qwen 3 VL 32B Instruct56.94
5Qwen 3 VL 8B Instruct56.49
6Qwen 3 VL 4B Instruct47.2
7Qwen 2.5 VL 32B Instruct43.8
8Qwen 2.5 VL 7B Instruct35.06
9InternVL3.5-8B26.53

Interactive version: theaggregate.ai/benchmark?slug=triviewbench-object-counting · How It Works · Data refreshed daily, snapshot 2026-09-29.