HoloCount - General Semantic Attribution: leaderboard
Counting the objects with a general semantic attribute: a subset of HoloCount (Meituan, 2026), a visual counting benchmark for multimodal LLMs. Models answer with a single integer, zero-shot; exact-match accuracy in percent, higher is better.
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 objects with a general semantic attribute (101 questions); higher is better). Source: arxiv.org. Saturation forecast: Not forecast. 30 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 3.1 Pro (Preview) | 87.1 | #94 |
| 2 | Gemini 3 Flash (Preview) | 86.1 | #144 |
| 3 | Qwen 3.5 27B | 83.2 | #359 |
| 4 | Qwen 3.5 35B A3B | 83.2 | #482 |
| 5 | Kimi K2.6 | 82.2 | #178 |
| 6 | Qwen 3.5 397B A17B | 80.2 | #246 |
| 7 | Qwen 3.5 122B A10B | 77.2 | #331 |
| 8 | Qwen 3.5 4B | 76.2 | #1175 |
| 9 | Qwen 3.5 9B | 75.2 | #803 |
| 10 | Gemini 2.5 Pro | 71.3 | #266 |
| 11 | GPT-5.5 | 68.3 | #49 |
| 12 | Kimi K2.5 | 68.3 | #256 |
| 13 | Claude Opus 4.8 | 68.3 | #56 |
| 14 | Qwen 2.5 VL 32B Instruct | 67.3 | #1053 |
| 15 | MiniMax-M3 | 66.3 | #269 |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=holocount-general-semantic-attribution · How It Works · Data refreshed daily, snapshot 2026-11-13.