Best Image Classification Models

Image classification is the task of assigning a label to an entire image from a set of categories, answering "what is this a picture of?". A model encodes the image, usually with a convolutional network like ResNet or EfficientNet or a vision transformer (ViT), and outputs a probability per class; the top score wins. CLIP-style models extend this to zero-shot classification: they embed the image and your candidate label texts in the same vector space and pick the closest match, so you can change categories without retraining. Accuracy is reported as top-1 or top-5 on the fixed label set. Classification differs from detection (which localizes objects) and tagging (which applies many loose keywords), and it powers quality inspection, content moderation, product categorization, and medical screening. This page lists 72 image classification models, including 33 open-weight options you can self-host; 61 of them run live in the Playground so you can test them on your own images.

It depends on your task and constraints. For fixed categories in production, a model fine-tuned on your own data typically beats any general-purpose model. Compare the image classification models on this page and try them on your own image to see which fits.

Yes. 33 of the 72 image classification models here are open-weight (for example Kimi K3, Qwen3.6 27B, and Qwen3.6 35B A3B), free to self-host under their licenses (Modified MIT, Apache 2.0, and Proprietary).

Yes. You can run 61 of them in the Roboflow Playground for free. Upload an image and compare the models' output side by side, no setup required.