Jan AI setup and tests: offline chat and HTML generation

This tutorial introduces Jan AI as an open source desktop app for Windows, macOS and Linux. The presenter explains how to download a model through its Hub and run models locally without registration or a credit card. Internet access is needed for the download; the downloaded model can then handle offline chat.

The tests cover a Persian arithmetic prompt, a five-person ordering puzzle and an HTML landing page. The presenter reports correct answers to the first two tasks and describes the generated page as basic. These examples show the results of a few prompts rather than establish general accuracy. The walkthrough also covers web search, thinking controls, and projects with files and instructions.

Hardware limits matter. Although the speaker says the selected model can work with less than 8 GB of RAM, he reports that memory use froze his camera image during recording. He also describes the tested setup as accepting text and Markdown files, without image or video analysis.

For developers, the video explains Jan's OpenAI-compatible API and using a running computer as an API server. It also discusses connecting Gemini through an API. The presenter treats Jan as useful for basic tasks rather than a full ChatGPT alternative.