Roboflow Vision Evals - Object Counting
Multimodal · 2026-07-22
The Roboflow Vision Evals counting task measures exact-match accuracy on 74 image questions that require counting visible objects.
Top models (higher is better)
| Model | Score |
|---|---|
| Gemini 3.6 Flash | 82.4 |
| Gemini 3.5 Flash | 81.1 |
| Muse Spark 1.1 | 75.7 |
| GPT-5.6 Sol | 73.0 |
| Gemini 3.1 Pro Preview | 71.6 |
| Gemini 3 Flash Preview | 67.6 |
| GPT-5.6 Terra | 67.6 |
| GPT-5.6 Luna | 66.2 |
| GPT-5.5 | 64.9 |
| Claude Fable 5 | 63.5 |
| GPT-5.4 Mini | 60.8 |
| Sonnet 5 | 56.8 |
| Qwen3.5-27B | 54.0 |
| Opus 4.8 | 52.7 |
| Gemini 2.5 Pro | 52.7 |
| Gemini 3.5 Flash-Lite | 52.7 |
| Kimi K2.6 | 50.0 |
| Qwen3.7-Plus | 50.0 |
| GLM 5V Turbo | 48.6 |
| Qwen3-VL-235B-A22B-Instruct | 47.3 |
| Kimi K3 | 46.0 |