General VQA on CountBench
97.35AccuracyGemini 3-Pro
Evaluation results
| Method | Links | |
|---|---|---|
| Gemini 3-Pro2026.02 | 97.35 | |
| ERNIE 5.02026.02 | 96.54 | |
| Bee-8BOpenness=fully open, Model Size=8B, Training Strategy=RL2025.10 | 93 | |
| Qwen3-VLmode=Thinking2026.02 | 92.67 | |
| Gemini 2.5-Pro2026.02 | 91 | |
| Bee-8BOpenness=fully open, Model Size=8B, Training Strategy=SFT2025.10 | 90.5 | |
| GPT-5tier=High2026.02 | 88.88 | |
| MolmoOpenness=fully open, Model Size=7B, Training Strategy=Distilled (-D)2025.10 | 84.8 | |
| Keye-VLOpenness=semi-open, Model Size=8B2025.10 | 78 | |
| Qwen2.5-VLOpenness=semi-open, Model Size=7B2025.10 | 74.1 |