Local AI

What this tool does

Local AI is a simple chat box. You type a question or a task, and a small language model writes an answer. The model is Llama 3.2 1B Instruct in a compressed 4-bit form. It runs on your own graphics card through your browser, not on a remote computer.

It handles one prompt at a time. It does not remember earlier questions, so put everything it needs in the same message. A small model is useful for short drafts, simple rewrites and ideas. It can make mistakes, so check facts before you use them.

This page does not open PDF files. To work with a document, try Summarize PDF or Ask PDF.

How to use it

1. Open the page and wait while it says Checking device capabilities.... If your device cannot run the model, you will see Not Supported and the reason.
2. On the first visit you see a notice that a model of about 800 MB will be downloaded. Make sure you have a good connection and enough free space.
3. Type your question in the box that says Ask something....
4. Select Download & Start. A progress bar and messages show the model loading. The button changes to Run once the model is saved on your device.
5. The answer appears in the Response box and grows word by word.
6. To stop, select Cancel. The words written so far stay on the screen.
7. Next time, select Run. The model is already saved, so you do not wait for the download again.

Privacy

On this page your prompt and the answer are handled by the model in your browser. They are not sent to a DocsBee server. The model files themselves are downloaded from the internet once and stored in your browser.

Tips

• Download the model on Wi-Fi if you can. It is large, and the first start is the slowest.
• Close other heavy tabs while it runs. The model uses your graphics memory, and older devices may be slow.
• Ask for one clear thing at a time. Short, specific prompts give better answers than long, vague ones.
• Do not trust answers about numbers, dates or laws without checking them.

Frequently asked questions