hjLabs AI Playground
Starting the on-device runtime…
Zero-shot classification lets you invent the categories at query time: type "billing, bug report, feature request, spam" and the model scores your text against each one. It works by reframing the problem as natural language inference with a BART model fine-tuned on MNLI, so no training data and no fine-tuning are involved. It is slower per call than a dedicated classifier but immeasurably faster to set up, and everything runs locally.
zero shot classificationbart mnli demotext classification without trainingcustom label classifiernli classification onlineintent classification freetopic labelling tool
Each label costs one forward pass, so ten labels take ten times as long as one. Up to about a dozen labels stays comfortably interactive on a GPU; beyond that consider training a real classifier.
Yes — switch off the single-label option and each label is scored independently, which is the right mode for tagging.
No. The model weights are downloaded from Hugging Face to your browser once, and every inference after that runs on your own GPU through WebGPU. Your text, images and audio are never sent anywhere — hjLabs.in has no inference server and no way to see your input.
Small models do. Recent mobile Chrome and Safari expose WebGPU, but memory is the binding limit — stay under roughly 200 MB of weights on Android, and expect iOS to reclaim the tab sooner. The catalog shows every download size up front so you can pick something that fits.
That depends on the individual model, not on this site. Each model page links to its Hugging Face repository, where the licence is stated — most of the catalog is Apache-2.0 or MIT, a few are research-only. We add nothing on top: hjLabs.in claims no rights over anything you produce here.