FAQ
Once your local deployment is up and running, the issues below are the most common. All answers are taken from the official Ollama FAQ; commands we have tested on our own machine are noted.
Check your issue first
Look it upWhat I'm running into is…
Click one to see.
What you'll be able to do after this lesson
- Move models to another drive
- Let other devices on the LAN access your Ollama
- Use only local models, free up GPU memory, upgrade Ollama
Where models are stored and how to move them to another drive
| OS | Default location |
|---|---|
| Windows | C:\Users\%username%\.ollama\models |
| macOS | ~/.ollama/models |
| Linux | /usr/share/ollama/.ollama/models |
To change the location: set the environment variable OLLAMA_MODELS to point to the new directory. All Ollama configuration is done via environment variables; setup methods for each system:
Windows: first quit Ollama from the taskbar; in "Settings" (Windows 11) or "Control Panel" (Windows 10), search for "environment variables", click "Edit your account's environment variables", and create or modify
OLLAMA_MODELS; after saving, reopen Ollama from the Start menu.Mac: when Ollama runs as an application, set it with
launchctl setenv, then restart the Ollama app.Linux (when running as a systemd service): run
systemctl edit ollama.service, add the lineEnvironment="OLLAMA_MODELS=新目录"under[Service], then after saving runsystemctl daemon-reloadandsystemctl restart ollama. Official reminder: on Linux, the new directory must be readable and writable by theollamauser.I know how to change
OLLAMA_MODELS
LAN access
By default, Ollama only listens on the local 127.0.0.1:11434, so other computers cannot connect. To let devices on the LAN access it, set OLLAMA_HOST to 0.0.0.0:11434, using the same setup method as above, then restart.
Example on Mac:
launchctl setenv OLLAMA_HOST "0.0.0.0:11434"Once enabled, any device on the same network can call your models — only do this on networks you trust.
- I know that LAN access requires changing
OLLAMA_HOST
Use only local models
Ollama also offers cloud models and web search. If you only want to use local models, you can disable the cloud features. Write this in ~/.ollama/server.json:
{
"disable_ollama_cloud": true
}Or set the environment variable OLLAMA_NO_CLOUD=1. After restarting Ollama, Ollama cloud disabled: true will appear in the logs. Once disabled, cloud models and web search are no longer available.
GPU memory stays occupied
By default, a model stays in memory for 5 minutes after use, so the next call is faster. To release it immediately:
ollama stop deepseek-r1:8bWhat you should see: the command produces no output; run ollama ps again and the model is no longer in the list. We tested this one on our own machine.
When calling via the API, you can control it with the keep_alive parameter: 0 unloads immediately after use, a negative value keeps it loaded indefinitely, and you can also specify durations like "10m" or "24h".
How to upgrade
- Mac and Windows: Ollama downloads updates automatically — click "Restart to update" in the taskbar or menu bar icon. You can also download the installer again from the official website.
- Linux: rerun the install command.
curl -fsSL https://ollama.com/install.sh | shFAQ
Will a locally deployed DeepSeek send my data out?
According to the official FAQ: Ollama runs locally, and when running locally they cannot see your prompts or data. Your data only passes through their services when using cloud models. If you only use local models, just disable the cloud features as described above in "Use only local models".
What is Ollama's default port?
- By default it only listens on the local
127.0.0.1:11434.
Checkpoint
- I can change
OLLAMA_MODELSandOLLAMA_HOST - I know how to use only local models
- I can use
ollama stopto free up GPU memory
Next lesson
Wondering whether the latest DeepSeek V4 can also be installed on your own computer? See Can DeepSeek V4 run locally. Once you have the basics covered, you can build a knowledge base of your own: Build your own knowledge base.