The Complete Guide to Ollama on a USB Drive (2026)
Ollama makes running local AI models dramatically easier than it used to be. And with the right setup, you can run Ollama from a USB drive, carrying your models and runtime between machines without installing anything on the host. This guide covers every step, including where DIY Ollama portable ends and where a preconfigured offline AI USB like PortableMind begins.
TL;DROllama can run from a USB drive with a portable setup: put the binary and model store on the drive, point OLLAMA_MODELS at it, and mind drive speed. PortableMind is the pre-built version of the same idea, with voice and vision included.
What is Ollama and why use it on a USB?
Ollama is an open-source tool that makes it easy to download and run large language models locally. It handles model management, API serving, and runtime configuration, essentially a package manager and server for local AI.
Running Ollama from a USB drive means your models and runtime travel with you. You plug the USB into any compatible machine and AI works, without installing Ollama on the host or downloading multi-gigabyte models every time you change computers.
Requirements for Ollama on USB
Hardware: USB 3.x flash drive (64 GB minimum, 128 GB recommended for multiple models). A USB SSD will be noticeably faster for larger models.
Host machine: Windows 10/11 or macOS with 8 GB+ RAM. Ollama uses your machine's CPU for inference; a GPU will speed things up but isn't required.
Model sizes: Llama 3.1 8B quantized = ~4.7 GB. Mistral 7B quantized = ~4.1 GB. Plan storage accordingly.
Setting up Ollama on a USB drive: step-by-step
Step 1: Format the USB drive as exFAT (works on both Windows and macOS).
Step 2: Create a folder structure on the USB: /ollama/bin/ for the Ollama executable, /ollama/models/ for model storage.
Step 3: Download the Ollama binary for your OS from the Ollama GitHub releases page.
Step 4: Set the OLLAMA_MODELS environment variable to point to your USB model directory before launching Ollama.
Step 5: On Windows, create a .bat launcher that sets the env var and starts Ollama. On macOS, create a .command script.
Step 6: Pull your first model: ollama pull llama3 (while connected to internet for the initial download).
Step 7: Disconnect internet, run your launcher script, and test with ollama run llama3.
DIY Ollama USB vs. PortableMind: when to use each
DIY Ollama USB is a legitimate option if you're technically comfortable, want free software, and enjoy customizing your setup. The trade-off: 2-4 hours of setup, per-OS launcher scripts, no official support, and no bundled voice/vision features.
PortableMind is a preconfigured offline AI USB that ships with Ollama-compatible models, a polished desktop app, voice mode, SCOUT Vision, and launchers for both Windows and macOS. $99 one-time. Plug in and go.
Most users are better served by PortableMind unless they specifically want to manage the model stack themselves.
- DIY: free, flexible, 2-4 hours setup, no support.
- PortableMind: $99, plug-and-run, voice + vision included, Windows + macOS launchers.
Ready to run AI offline?
PortableMind is the plug-and-run offline AI USB. Go Digital for $69 and flash your own drive, or get the $99 v1.8.0 USB (voice, SCOUT Vision field modes, multi-modal chat, in-app Model Manager) or MAX-SPEED for power users. No internet, no subscription.
Conclusion
Running Ollama from a USB drive is a real, practical setup for technical users who want portable offline AI. If you'd rather skip the setup and just have it work, PortableMind is the plug-and-run version of the same idea, models preloaded, voice and vision included.
See the PortableMind Ollama USB setup guide →Frequently asked questions
Long-tail answers for the search queries around this topic.
- Can you run Ollama from a USB drive?
- Yes. Set the OLLAMA_MODELS env variable to a USB directory and run the Ollama binary from the USB. Works on Windows and macOS.
- What USB size do I need for Ollama?
- 64 GB minimum for one model. 128 GB+ recommended if you want multiple models. A USB SSD will be faster than flash for large models.
- Is Ollama portable?
- It can be made portable with some setup. PortableMind ships as a ready-made portable Ollama-compatible setup with no configuration required.
- What is the best model to run on Ollama from a USB?
- Llama 3.1 8B (quantized) or Mistral 7B are good starting points for most hardware. Both run well on 8 GB RAM.
- Does Ollama on USB work without internet?
- Yes. After initial model download, Ollama runs fully offline. PortableMind ships with models preloaded so no download is needed at all.
- How do I move Ollama models to a USB?
- Copy the models directory to the USB, then set OLLAMA_MODELS to the USB path before launching Ollama. On Windows this is done via a .bat script; on macOS via a .command script.
Carson Dresser created the first offline AI USB. He builds from South Florida, where hurricane season makes losing power a lived reality rather than a thought experiment, and he watched cloud AI fail people at the worst possible moments: locked accounts, surprise price hikes, assistants that vanish with the signal. In 2025 he decided intelligence should be something you own, and shipped it. Working alone, he engineered the firmware, tuned the local models, wrote the desktop app, and built the over-the-air update system that delivers cryptographically signed payloads to drives already in the field. He also designed the product, built this website, and runs the store, the kind of range that usually takes a company. Before PortableMind he built ClipStitcher and a stack of practical automations for creators and small operators. Every order that ships and every support email still goes through Carson personally.