Supported models

There are three ways you can deploy models with Inference:

1. Deploy a model hosted on Roboflow onto your device.
2. Upload suported model weights to Roboflow, then deploy to your device.
3. Deploy a foundation model out-of-the-box.

The pages linked above will walk you through each of these options.

Supported Custom Models¶

The following custom models are supported:

• YOLOv5 (Segmentation, Object Detection)
• YOLOv7 (Segmentation)
• YOLOv8 (Classification, Segmentation, Object Detection)
• YOLO-NAS (Coming soon)

Supported Foundation Models¶

• CLIP
• CogVLM
• DocTR
• Grounding DINO
• L2CS-Net
• Segment Anything
• YOLOWorld