Ollama

Updated Jun 13, 2026 ·

Overview

Ollama is a tool for running AI models on your own computer.

It works on macOS, Linux, and Windows. Unlike LM Studio, Ollama does not have a built-in graphical interface. You mainly use it from the terminal.

Install Ollama

Download Ollama from the official website.

Go to: https://ollama.com

Choose the version for your system.

• macOS
• Linux
• Windows

After installing, start Ollama.

• On macOS, you may see it in the status bar.
• On Windows, you may see it in the system tray.

Verify the installation by checking the version.

ollama --version

Output:

ollama version is 0.30.8

If the command is not found, restart your terminal or restart your system. Also make sure Ollama was installed and started properly.

Cloud Models in Ollama

Some Ollama models are cloud models. These do not run locally.

For example, if you run a cloud model like this:

ollama run minimax-m3:cloud

You may be asked to sign in.

Example output:

To use minimax-m3:cloud, please sign in.

Navigate to:
https://ollama.com/connect

After signing in, Ollama will ask you to pair your device with your account.

You can check your models with this command:

ollama list

Output:

NAME                ID              SIZE    MODIFIED
minimax-m3:cloud    d03a959f45c0    -       About a minute ago

If the size shows -, it means the model is not stored locally. It is a cloud model.

Pull a Model Without Running It

You can download a model without starting a chat session.

In the example below, gemma4:e4b is downloaded only.

ollama pull gemma4:e4b

After that, verify it with:

ollama list

Output:

NAME         ID              SIZE      MODIFIED
gemma4:e4b    abc123example   3.3 GB    1 minute ago

Use ollama pull when you want to prepare the model first and run it later.

Running a Local Model

To run a local model, use ollama run followed by the model name.

In the example below, gemma4:e4b is the model name and tag.

ollama run gemma4:e4b

If you haven't pulled the model yet, the first run downloads it.

After the download finishes, Ollama starts the model in the terminal.

Output:

>>>

This means Ollama is ready for your message.

Basic Chat Commands

Inside the Ollama chat, you can use slash commands.

Command	Description
/?	Show available help and commands
/bye	Exit the current Ollama session
/show info	Display information about the current model
/show parameters	Display the active model parameters and settings
/show system	Display the current system prompt

In the example below, /show info displays details about the running model.

>>> /show info

  Model
    architecture        gemma4    
    parameters          8.0B      
    context length      131072    
    embedding length    2560      
    quantization        Q4_K_M    
    requires            0.20.0    

  Capabilities
    completion    
    vision        
    audio         
    tools         
    thinking      

  Parameters
    top_p          0.95    
    temperature    1       
    top_k          64      

  License
    Apache License               
    Version 2.0, January 2004    
    ...          

NOTE: The context length shown here is the maximum supported by the model. It may not always be the active context length for the current session.

Sending a Multi-line Message

By default, pressing Enter sends your message.

To write multiple lines, use triple double quotes.

"""
Hi
How are you?
This is a multi-line message.
"""

This is useful when you want to send longer prompts.

Sending an Image to a Vision Model

Some models can understand both text and images.

For example, Gemma 4 has vision support, depending on the version you use.

Here, the prompt comes first, and the image path comes last.

What's on this image? ./images/sample.jpg

Output:

The image shows a person in a room with a recording setup.

The image path must be placed at the end of the prompt. Ollama uses the path to attach the image to your message.

Check Model Details

You can inspect the current model with /show.

/show info

This can show details like:

• Model family
• Parameter size
• Context length
• Quantization
• Capabilities
• Active parameters

You can also check the current parameters:

/show parameters

Output:

Model defined parameters:
  top_p                          0.95
  temperature                    1
  top_k                          64

These are the default settings applied by Ollama for the model.

Ollama Model Catalog

You can browse available models on the Ollama website.

https://ollama.com/search

You can find models like:

• Gemma
• Llama
• Mistral
• DeepSeek

For vision models, you can search for models with image support.

https://ollama.com/search?c=vision

The model page usually shows the command needed to run the model.

This is the simplest way to find the correct model name and tag.

Ollama Model Tags

Ollama models use tags to identify specific versions.

Example:

gemma4:e4b

In this example:

• gemma4 is the model name
• e4b is the tag
• The tag usually describes the model size or version

Some tags also refer to quantized versions.

Quantized models use less disk space and need less hardware. This is why most Ollama models are already quantized by default.

Ollama does not include a GUI

Ollama mainly runs from the terminal.

If you want a graphical interface, you can use:

• LM Studio
• Open WebUI

LM Studio is easier if you want a simple desktop app.

Open WebUI is a browser-based interface, but it usually needs Docker or more setup.

info

Ollama is best if you are comfortable with the terminal and want a lightweight way to run local models.