Essential Ollama CLI Commands

Streamline your local AI model workflow with the Ollama CLI. If you have experience with Docker, many of these commands will feel instantly familiar.

Note

Most Ollama commands mirror Docker syntax for running, listing, and managing images—making it easy to get started if you’ve used Docker before.

Quick Reference

Command	Purpose	Example
ollama run	Pull (if needed) and start a model	`ollama run MODEL_NAME`
ollama stop	Terminate a background model	`ollama stop MODEL_NAME`
ollama list	Display all local models	`ollama list`
ollama rm	Remove a model to free disk space	`ollama rm MODEL_NAME`
ollama pull	Download a model without launching it	`ollama pull MODEL_NAME`
ollama show	Inspect model metadata (quantization, etc.)	`ollama show MODEL_NAME`
ollama ps	List currently running models	`ollama ps`

ollama run

Start a model session. If the specified model isn’t on your machine, Ollama will:

Download its manifest and layers
Verify the SHA-256 digests
Save the manifest locally
Launch an interactive chat interface

$ ollama run MODEL_NAME
pulling manifest
pulling 633fc5be925f... 100%
pulling fa8235e5b48f... 100%
pulling 542b217f179c... 100%
pulling 8de1baf1db0... 100%
pulling 23291dc44752... 100%
verifying sha256 digest
writing manifest
success

ollama stop

When you run a model in the background (e.g., via & on Unix), stop it gracefully with:

$ ollama stop MODEL_NAME

This frees up GPU/CPU resources immediately.

ollama list

List all models stored on your local host, along with their IDs, sizes, and last modification times. Use this to identify large or outdated models:

$ ollama list
NAME             ID              SIZE    MODIFIED       
phi3:latest      4f2229297938   2.2 GB  20 hours ago    
qwen:latest      d53d04290064   2.3 GB  12 days ago     
mistral:latest   f974a74358d6   4.1 GB  13 days ago     
llama3.2:latest  a80c4f17acd5   2.0 GB  13 days ago

ollama rm

Remove a model from your system to reclaim disk space. You can always re-pull it later.

$ ollama rm MODEL_NAME
deleted 'MODEL_NAME'

ollama pull

If you want to prefetch a model without starting it immediately, use:

$ ollama pull MODEL_NAME
pulling manifest
pulling dde5aa3fc5ff... 100%   2.0 GB
pulling 966de95ca8a6... 100%   1.4 KB
pulling fcc5a6bec9da... 100%   7.7 KB
pulling a70ff7e570d9... 100%   6.0 KB
pulling 56bb8bd477a5... 100%   96 B
pulling 34bb5ab01051... 100%   561 B
verifying sha256 digest
writing manifest
success

After pulling, a subsequent ollama run MODEL_NAME will skip downloads.

ollama show

Inspect model metadata—such as license details and quantization format—without running it:

$ ollama show llama3.2
License
LLAMA 3.2 COMMUNITY LICENSE AGREEMENT
Llama 3.2 Version Release Date: September 25, 2024
quantization    Q4_K_M
...

You can also view model details on the Ollama website or from within a running session via the /show command.

ollama ps

Similar to docker ps, this command lists all active model processes, their resource consumption, and remaining runtime:

$ ollama ps
NAME             ID              SIZE    PROCESSOR    UNTIL
phi3:latest      4f2229297938    6.6 GB  100% GPU     4 minutes from now
llama3.2:latest  a80c4f17acd5    4.0 GB  100% GPU     4 minutes from now

Warning

Keep an eye on running models. Stopping unused ones prevents unnecessary GPU/CPU usage and conserves system memory.

These core commands will help you manage Ollama models effectively. Happy modeling!

Watch Video

Watch video content