HoloCount - Visual Distractor Illusion: leaderboard

Metric: Accuracy (%; exact-match accuracy of the single integer count a multimodal model returns zero-shot for an image and a counting question, with a strict integer-only system prompt; counting real objects among look-alike distractors such as plastic fruit (200 questions); higher is better). Source: arxiv.org. Saturation forecast: Around March 2027. 30 models tracked.

Top models

#ModelScore
1Qwen 3.5 397B A17B80
2Kimi K2.678.5
3Qwen 3.5 122B A10B78
4Qwen 3.5 35B A3B77.5
5Qwen 3.5 9B76.5
6GPT-5.576
7Kimi K2.573.5
8Claude Opus 4.873.5
9GPT-4.173
10Qwen 3.5 27B73
11Gemini 2.5 Pro71
12Qwen 3 VL 32B Instruct71
13Claude Opus 4.770.5
14Qwen 3.5 4B70.5
15MiniMax-M369

Interactive version: theaggregate.ai/benchmark?slug=holocount-visual-distractor-illusion · How It Works · Data refreshed daily, snapshot 2026-09-29.