Benchmark 是了解硬件平台网络模型运行速度的最佳途径。仅供社区参考，不代表商业交付最终性能。
- Pulsar2 2.5
| Models | Input Size | Inference Time(ms)@3.2T |
|---|---|---|
| Inceptionv1 | 224 | 1.795 |
| Inceptionv3 | 299 | 7.065 |
| MobileNetv1 | 224 | 0.901 |
| MobileNetv2 | 224 | 0.923 |
| SqueezeNetv11 | 227 | 0.824 |
| SqueezeNetv10 | 227 | 1.097 |
| ResNet18 | 224 | 2.540 |
| ResNet50 | 224 | 5.616 |
| VGG16 | 224 | 32.404 |
| YOLOv3 | 608 | 65.117 |
| YOLOv5s | 640 | 12.707 |
| YOLOv6s | 640 | 13.708 |
| YOLOv7-Tiny | 640 | 12.508 |
| YOLOv8s | 640 | 16.912 |
| YOLO11s | 640 | 15.933 |
| YOLOX_s | 640 | 15.425 |
| PPYOLOE+_s | 640 | 14.885 |
| Deit_t | 224 | 3.350 |
| Swin_t | 224 | 14.459 |
| ViT_b | 224 | 26.711 |