Live local machine learning · Free-first

Try a real AI model without a paid API.

Explore a real English text-sentiment model running on your own device. This is a classification experiment, not a conversational chatbot, a psychological assessment, or a substitute for human judgment.

Know before you load: Clicking the download button loads Transformers.js from jsDelivr and quantized DistilBERT weights from Hugging Face (approximately 68 MB plus runtime files). Downloads consume data, require a stable connection, and may fail on low-memory devices. After loading, inference runs locally through WebAssembly. The software publisher and model host receive ordinary asset requests; your analysis text is processed by the browser worker and is not sent to an AI University inference server.

Step 1 · Load the open model

Model not loaded. Confirm download consent to begin.

Model: Xenova/distilbert-base-uncased-finetuned-sst-2-english ↗. Model's base license: Apache 2.0. JavaScript engine: Transformers.js 3.8.1 ↗.

Step 2 · Classify a short English sentence

Text to analyze (maximum 350 characters)

Try these examples and discuss where the model gets confused.

This binary classifier returns POSITIVE or NEGATIVE with a model score. It does not understand nuanced emotions, satire, health risks, truthful claims, or the full meaning of a statement.

Step 3 · Keep the human in control

Make your own first classification before viewing the model's result. Compare both opinions, change one word, run again, and record where they disagree. This is a teaching example of AI-assisted evaluation, not a guarantee of accuracy.

Compare other open models ↗ · Return to Free AI Lab ↗