Managing Ollama Models

Introduction

Once you’ve installed Ollama, the next step is downloading models, organizing them, and creating custom variants when needed. You can handle everything from the command line efficiently. This article covers the essential commands and walks you through creating your own model files.

Ollama Models Management - In A Nutshell

Ollama stores all models locally in ~/.ollama/models. Use ollama pull name to download a model. Use ollama list to see what’s installed. Use ollama rm name to delete one. If you want to run a model with a custom system prompt or temperature setting, create your own Modelfile.

Key Terms and Components

Term	Definition
pull	Downloads a model from the Ollama library
list	Shows all locally installed models
rm	Deletes a model from your system
run	Launches a model for chat
Modelfile	A file that specifies quantization, system prompt, and parameters

When It Matters

When Do You Need to Manage Models?

Once you start working with multiple models, disk space adds up quickly. Some models you’ll test once and never use again. Others you’ll want to customize with specific system prompts for customer support, coding, or creative writing tasks. Proper management keeps things organized and frees up storage.

Common Commands at a Glance

# Download a model
ollama pull llama3

# List all local models
ollama list

# Start a model
ollama run llama3

# Show currently running processes
ollama ps

# Delete a model
ollama rm llama3

The ollama ps command is particularly useful for checking which models are currently loaded in memory and consuming RAM.

Defining Your Own Model

A Modelfile sets parameters like temperature, system prompt, and quantization. Here’s a simple example:

FROM llama3

SYSTEM "You are a helpful assistant. Respond with precision and clarity."

PARAMETER temperature 0.7

Save this as MeinModell and create the model with:

ollama create meinmodell -f MeinModell

Then launch it using ollama run meinmodell.

Storage, Cost, and Security Considerations

Each downloaded model takes up disk space. With many models, you can quickly accumulate dozens of gigabytes. Regular cleanup with ollama rm helps preserve space. Costs are zero since Ollama and most open-source models are free. From a security standpoint, all models and configurations stay on your machine.

Key Takeaways on Model Management

• ollama pull downloads models, ollama rm removes them.
• ollama list shows installed models, ollama ps shows currently running ones.
• Modelfiles let you create custom configurations with system prompts and parameters.
• Regularly deleting unused models saves disk space.

Learn more about installation in Installing Ollama and the API in Using the Ollama API.

FAQ - Common Questions About Model Management

Can I have multiple versions of the same model?

Yes. Each version takes up its own storage. You can create them with individual Modelfiles.

Where are models stored on disk?

By default, ~/.ollama/models on Linux and macOS, or in the Ollama user directory on Windows.

How much storage does a model use?

A quantized 7B model takes roughly 4 to 5 GB, a 13B model about 8 to 9 GB, depending on quantization level.

Can I create Modelfiles from GGUF files?

Yes. Use FROM ./path/to/model.gguf in your Modelfile and import it with ollama create.

Tools and Further Resources

Beyond the Ollama CLI, graphical tools like Open WebUI and LM Studio offer a more visual way to manage models. For automated workflows, the CLI commands and API are practical options.

Sources

• Ollama Model Management
• Ollama Modelfile Reference