Ollama on USB FAQ
PortableMind ships with Ollama ready to run offline. These FAQs explain models, caching, and what to do if Ollama tries to download.
Basics
Does Ollama need internet here?
No. Models are preloaded; keep Wi-Fi off after launch.
Can I add models?
Yes—download on a trusted network once, then validate offline.
Where are models stored?
On the USB in the preloaded models directory; keep free space available.
Performance
Which model for low-spec?
Use the smaller quantized models in the low-spec preset to keep responses fast.
How to reduce lag?
Shorten prompts, reduce stream speed, and close heavy apps.
Should I stay plugged in?
For heavier models, yes. Battery is fine for low-spec tasks.
Troubleshooting
Model tries to download
Cancel, pick a preloaded model, and update later on a trusted network.
Ollama not responding
Check the terminal is running, refresh localhost, and ensure VPN/proxy isn’t blocking.
Blank UI
Wait 40 seconds, clear cache for localhost, and relaunch if needed.
Management
How to track available models?
Keep a manifest file on the USB listing model names and presets.
Can I clone to another drive?
Yes. Clone carefully, verify paths, and keep the original USB as backup.
Do updates replace models?
Only if specified. Back up before updating and revalidate models afterward.
More resources
Related guides
Best offline AI models for Mac: what runs well on MacBooks
Model guidance for macOS: what runs well on MacBooks and how to stay stable without cloud dependencies.
Best offline AI models for Windows: fast vs smart presets
Model guidance for Windows laptops: choose lightweight models for speed or bigger ones for reasoning, with real-world tips.
Offline AI USB FAQ: quick answers
Fast answers to common PortableMind offline AI USB questions about setup, privacy, models, and troubleshooting.