Multimodal Perception on ZoomBench
59.29AccuracyGemini-3-Flash
Evaluation Results
| Method | Links | |
|---|---|---|
| Gemini-3-FlashType=Closed-Source2026.02 | 59.29 | |
| ZwZ-8Bnumber of parameters=8B, Type=Our Model2026.02 | 58.11 | |
| Kimi-K2.5 (1T)number of parameters=1T, Type=Open-Source2026.02 | 56.33 | |
| ZwZ-4Bnumber of parameters=4B, Type=Our Model2026.02 | 55.74 | |
| ZwZ-7Bnumber of parameters=7B, Type=Our Model2026.02 | 55.62 | |
| GLM-4.5V (108B)number of parameters=108B, Type=Open-Source2026.02 | 49.23 | |
| Qwen3-VL-235B-A22Bnumber of parameters=235B, Type=Open-Source2026.02 | 49.11 | |
| GPT-5.1Type=Closed-Source2026.02 | 47.22 | |
| MiMo-VL-7B-RLnumber of parameters=7B, Type=Open-Source2026.02 | 45.09 | |
| MiniCPM-V-4.5 (9B)number of parameters=9B, Type=Open-Source2026.02 | 42.6 | |
| Qwen2.5-VL-7Bnumber of parameters=7B, Type=Open-Source2026.02 | 42.49 | |
| Qwen3-VL-4Bnumber of parameters=4B, Type=Open-Source2026.02 | 40.24 | |
| Qwen3-VL-8Bnumber of parameters=8B, Type=Open-Source2026.02 | 37.87 |