Here is the step-by-step guide to set up a private AI model on a USB drive: 1. Prepare the USB Drive Insert USB: Plug a USB drive (at least 8GB recommended) into your computer. Format Drive: Right-click the drive in File Explorer, select Format, set the File System to exFAT, check Quick Format, and click Start. 2. Download Llamafile Search GitHub: Open Google, search for ⁠llamafile github⁠, and click the official GitHub link. Get Executable: Go to the Releases section on the right, click the latest release (e.g., ⁠v0.10.5⁠), and download the main ⁠llamafile⁠ binary. Rename File: Once downloaded, rename the file extension to end with ⁠.exe⁠ (e.g., ⁠llamafile-0.10.5.exe⁠). Copy to USB: Move or copy this ⁠.exe⁠ file directly into your USB drive. 3. Download an AI Model (GGUF) Navigate to Hugging Face: Go to ⁠huggingface.co⁠, click Models, and filter by the GGUF library option on the left. Search Model: Search for a lightweight model, such as ⁠qween⁠ or ⁠qwen3⁠. Download ⁠.gguf⁠: Click on the model, open Files and versions, and download the ⁠.gguf⁠ quantization file (around 2GB to 4GB). Copy to USB: Move the downloaded ⁠.gguf⁠ file onto your USB drive. 4. Create the Startup Script Open Notepad: Launch a new Notepad window Enter Command: Type the following code, replacing the file names with your exact downloaded file names: .\llamafile-0.10.5.exe --server --model qwen3-4b-thinking-2507.Q4_K_M.gguf Save File: Select File > Save As, name the file ⁠start.bat⁠, set "Save as type" to All Files, and save it inside your USB drive. 5. Launch and Run the AI Run Script: Double-click ⁠start.bat⁠ inside your USB drive to open the command window. Access UI: Look for the local address listed at the bottom (usually ⁠[http://127.0.0.1:8080](http://127.0.0.1:8080)⁠) and open it in your web browser. Start Chatting: You now have a fully functional, offline, and private AI model running directly off your USB stick. - @futureshiftt