Roboflow + YOLOv8
Export Roboflow datasets for YOLOv8, train detection, segmentation or keypoint models with Ultralytics, then upload the weights to run them on Roboflow.
How to connect YOLOv8 to Roboflow
6 stepsLabel and generate a version
Annotate boxes, polygons or keypoints in Roboflow, then generate a dataset version.
Download in YOLOv8 format
Call project.version(n).download("yolov8") or choose YOLOv8 on export. Roboflow writes the data.yaml with your class names.
Pin the Ultralytics version
Install ultralytics==8.0.196 before training. Our docs require that version for YOLOv8 weight upload.
Train the model
Run yolo task=detect mode=train with data pointing at the data.yaml, or open the matching YOLOv8 notebook in roboflow/notebooks.
Upload the weights
Call version.deploy(model_type="yolov8", model_path="runs/detect/train/"), using yolov8-seg for segmentation runs. The SDK reads weights/best.pt.
Run the model
Call it through the Serverless Hosted API, Workflows or a self-hosted Inference server with your workspace and model ID.
What teams build with YOLOv8
Test YOLOv8 before you label anything
Run the yolov8n-640 or yolov8x-seg-640 COCO checkpoints on your images through the API with no training.
Serve keypoint and segmentation models
Upload YOLOv8 pose or segmentation weights and call them over HTTP or in a Workflow.
Move from YOLOv8 to an in-app model
Train RF-DETR or YOLO11 in Roboflow on the same dataset version and compare with your YOLOv8 run.
Questions
No. Our docs list YOLOv8 for pretrained aliases and weight upload only. RF-DETR, YOLO26 and YOLO11 train in the app.
Uploaded YOLOv8 weights must be trained on ultralytics==8.0.196. With a newer version installed, the SDK warns about the mismatch before uploading.
Yes. YOLOv8 uses the YOLOv5 PyTorch TXT layout: one txt per image with normalized class, center and size values, plus a data.yaml.
YOLOv8 is AGPL-3.0. On Roboflow cloud a commercial license is included for every user, including the free plan. Commercial self-hosting needs our Enterprise license add-on.
Why are companies choosing Roboflow?
The end-to-end platform for vision AI: create datasets, train models and deploy to production.
Flexible deployment modes
Move from a scalable cloud API to edge hardware like NVIDIA Jetson without managing complex dependencies.
Integrated workflow builder
Combine multiple AI models, custom business logic and third-party integrations into one production-ready pipeline in a low-code editor.
Model monitoring and alerts
Dashboards track confidence, inference time and request volume for your deployed models, and email alerts tell your team when confidence suddenly drops.
