Ask questions about a small dataset in this browser.

Paste a small dataset and ask a question. A language model runs inside this browser to produce the answer, so there is no API key and your data is not uploaded.

First run downloads a model. The language model is fetched from Hugging Face the first time you run this — about 276 MB, so use Wi-Fi — and cached in your browser afterwards. The tool code is loaded from the jsDelivr CDN, and your data is not sent anywhere.

Paste a small dataset

Use comma-separated values (CSV) or a short list. Your data stays in this browser.

Checking whether this browser can run the model…

The first run downloads about 500 MB and saves it in this browser for later visits.

Example answer · load the model to ask your own questionThe file appears to list transport stops. One row has no stop name or place, two rows are marked inactive, and two update dates are several years old. “Bank of Engalnd” may be a misspelling of “Bank of England”.

How it works. The language model reads the data and question in a browser worker, then writes the answer a little at a time. The page does not send the dataset to an API.

Technical note: WebLLM runs the Qwen2.5-0.5B model through WebGPU, the browser feature that uses the device's graphics processor.