Pretrained Model Aliases
Built-in COCO-pretrained model aliases you can run with Roboflow Inference without training a model.
Roboflow Inference ships with a set of pretrained models trained on the COCO dataset and exposed under short aliases such as rfdetr-small or yolov8n-640.
Aliases are convenience shortcuts: you do not need to know the underlying Roboflow project and version, and you can pass the alias anywhere a model ID is accepted.
Code sample
Get your API Key
Create a Roboflow account, find your key on the Roboflow API settings page and make it available to your shell:
export ROBOFLOW_API_KEY="your-key-here"Run the model
Pass the alias as model_id:
import os
from inference_sdk import InferenceHTTPClient, InferenceConfiguration
client = InferenceHTTPClient(
api_url="https://serverless.roboflow.com",
api_key=os.environ["ROBOFLOW_API_KEY"],
).configure(InferenceConfiguration(api_key_transport="header"))
results = client.infer(
"https://media.roboflow.com/inference/people-walking.jpg",
model_id="rfdetr-small",
)
print(results["predictions"])Install the dependencies
Install the Inference package to load weights into your own process:
pip install inferenceAvailable aliases
Click "Test in browser" to try a model on Roboflow Universe before deploying it.
RF-DETR
Model page: RF-DETR
YOLO26
Model page: YOLO26
YOLO11
Model page: YOLO11
YOLOv10
Model page: YOLOv10
YOLOv8
Model page: YOLOv8
YOLO-NAS
Model page: YOLO-NAS
Related pages
Supported Models - the full catalog of models you can deploy with Roboflow.
Model IDs - how to reference your own trained models.
Find a Model on Universe - 50,000+ community-trained models you can run the same way.
Last updated
Was this helpful?