Supported Models

Roboflow Inference supports a wide range of models for computer vision tasks. You can run fine-tuned models trained on your own data, pre-trained models from Roboflow Universe, and foundation models for zero-shot tasks.

Quick Start

Using a model with Inference is straightforward:

from inference import get_model

model = get_model(model_id="rfdetr-small")

results = model.infer("https://media.roboflow.com/inference/people-walking.jpg")
Pre-Trained Model Aliases

Inference provides convenient aliases for common pre-trained models. See the Pre-Trained Aliases page for a full list of available model IDs.

Model Categories

Fine-Tuned Models

Fine-tuned models are trained on your own data using Roboflow or external training pipelines. Inference supports the following fine-tuned model architectures:

ArchitectureTasks
RF-DETRObject Detection
YOLO26Object Detection, Instance Segmentation, Keypoint Detection
YOLOv11Object Detection, Image Segmentation, Keypoint Detection
YOLOv10Object Detection
YOLOv9Object Detection
YOLOv8Object Detection, Classification, Image Segmentation, Keypoint Detection
YOLOv7Classification
YOLOv5Object Detection, Classification, Image Segmentation
YOLO-NASObject Detection, Keypoint Detection

Foundation Models

Foundation models are pre-trained on large datasets and can be used out of the box for a variety of tasks without additional training. See the Foundation Models overview for more details.

ModelUse Case
CLIPImage classification, embedding similarity
Grounding DINOZero-shot object detection
SAM 3Open-vocabulary segmentation
SAM 2Interactive image segmentation
SAMImage segmentation
Florence-2Object detection, captioning, OCR, and more
YOLO-WorldZero-shot object detection
PaliGemmaVQA, object detection, segmentation
Qwen 3.5-VLMultimodal understanding, VQA
SmolVLM2VQA, document OCR
Moondream2Image captioning, VQA
DocTROCR
TrOCROCR (single-line text)
GLM-OCROCR
OWLv2Few-shot object detection
Perception EncoderImage/text embeddings
Depth EstimationDepth maps
Gaze DetectionGaze estimation

Universe Models

Roboflow Universe hosts over 50,000 pre-trained models shared by the community. You can use any of these models with Inference.

Pre-Trained Model Aliases

Inference provides convenient IDs for common pre-trained models that do not require an API key. See the full list on the Pre-Trained Aliases page.