How to Update Ollama Without Losing Your Models

How to Update Ollama Without Losing Your Models: A FutureFormDigital Guide

Keeping your local AI models running smoothly means keeping your Ollama installation and its models up-to-date. It sounds simple, but sometimes the process can feel a bit murky. Don’t worry, at FutureFormDigital, we’re all about helping you build resilient, independent digital workflows. Today, we’re diving deep into how to update Ollama and its models without any drama, ensuring your AI experiments stay on track.

Why Bother Updating Ollama?

Think of Ollama as the engine for your local AI models. Like any engine, it gets upgrades! Newer versions often bring:

• New Features: Support for exciting new model architectures.
• Bug Fixes: Smoother operation and fewer surprises.
• Improved Performance: Faster model inference.
• Compatibility: Crucial for running the latest models like Llama 4 or Gemma 4, which might require a recent Ollama version to function at all.

Staying current is key to unlocking the full potential of the ever-evolving world of AI.

Understanding the Two Types of Updates

Before we jump into commands, let’s clarify what we’re actually updating. This is where many people get confused:

1. The Ollama Application: This is the core software that runs on your machine, serving the API and managing your models. Updating this is like getting a new version of your operating system.
2. Your Models: These are the actual AI models you’ve downloaded (like Llama 3.1, Mistral, Gemma, etc.). They need to be updated separately.

Let’s tackle each one.

Step 1: Update the Ollama Application

The method for updating the Ollama application itself varies slightly depending on how you installed it.

Checking Your Current Version

First off, know where you stand. You can check your Ollama version with:

ollama --version

Or via the API:

curl http://localhost:11434/api/version

Compare this to the latest releases on GitHub to see if you’re behind.

Updating on Different Operating Systems

• macOS:
  
  Ollama on macOS usually updates itself in the background. You’ll get a notification in your menu bar. Simply click the Ollama icon, select “Restart to Update,” and you’re good to go. If you installed via Homebrew, you can run:

  
  

  brew upgrade ollama

  
  

• Windows:
  
  Similar to macOS, Ollama for Windows checks for updates automatically. You’ll be prompted when a new version is ready. The easiest way is to download the latest installer from ollama.com/download and run it. It will upgrade your existing installation in place, preserving all your downloaded models automatically.

  
  

• Linux:
  
  Ollama on Linux doesn’t auto-update. You’ll need to re-run the official install script. This script is smart enough to update your existing installation without touching your models or configuration.

  
  

  curl -fsSL https://ollama.com/install.sh | sh

  
  

  After the script finishes, it’s a good practice to restart the Ollama service:

  
  

  sudo systemctl restart ollama

  
  

  You can check its status with sudo systemctl status ollama and view recent logs using journalctl -u ollama -n 50.

  
  

• Docker:
  
  If you’re running Ollama via Docker, updating involves pulling the new image and recreating the container. The key to not losing your models here is using Docker volumes.

  1. Pull the latest image:
     

     docker pull ollama/ollama:latest

     
     
  2. If using Docker Compose:
     

     docker compose pull
     docker compose up -d

     
     
  3. If using docker run directly:
     
     You’ll stop, remove, and then recreate the container, ensuring you map the same volume where your models are stored (usually /root/.ollama inside the container).
     

     docker stop ollama
     docker rm ollama
     docker run -d --gpus=all -v ollama:/root/.ollama -p 11434:11434 --name ollama ollama/ollama

     
     

     Crucially, the -v ollama:/root/.ollama flag tells Docker to use the named volume ollama for storing Ollama’s data, which includes your models. As long as you use the same volume name when recreating the container, your models will be safe.

     
     

Step 2: Update Your Ollama Models

Now that your Ollama application is up-to-date, let’s ensure your AI models are too. Ollama models are updated by simply pulling them again.

Updating a Single Model

If you only want the latest version of a specific model, like Llama 3.1, just run:

ollama pull llama3.1

Replace llama3.1 with the name of any model you have installed and want to update. To see all your installed models, use:

ollama list

Updating All Models at Once

There isn’t a single ollama update --all command, but you can achieve this with a clever one-liner.

• For Linux/macOS (Bash):

  
  

  ollama list | tail -n +2 | awk '{print $1}' | xargs -I {} ollama pull {}

  
  

  This command lists your models, skips the header, extracts each model name, and then runs ollama pull for each one.

  
  

• For Windows (PowerShell):

  
  

  ollama list | Select-Object -Skip 1 | ForEach-Object { $model = ($_ -split 's+')[0]; ollama pull $model }

  
  

  This does the same thing on Windows.

  
  

  
  
  Heads up: These scripts can take a while if you have many large models. It’s best to run them when you’re not actively using Ollama for other tasks.
  
  
  

When you pull a model that has been updated, Ollama checks for new layers or changes. It will only download what’s new, so you won’t re-download the entire model if only a small part has changed. Plus, these downloads are resumable if they get interrupted.

Troubleshooting Common Update Issues

Sometimes, things don’t go as planned. Here are a few common hiccups and how to fix them:

• Ollama Service Not Starting (Linux): If sudo systemctl restart ollama fails, check the logs: journalctl -u ollama -n 100. Sometimes, old library files can conflict. Try removing them and re-running the install script:

  
  

  sudo rm -rf /usr/lib/ollama
  curl -fsSL https://ollama.com/install.sh | sh
  sudo systemctl restart ollama

  
  

• Model Behaving Oddly After Update: The way Ollama interacts with models (its template format) can sometimes change between application versions. If a model acts strange, try re-pulling it: ollama pull modelname.

  
  

• Not Enough Disk Space: Models can be large! Check where they’re stored (~/.ollama/models on Linux/macOS, C:\Users\YourName\.ollama\models on Windows). Free up space by removing models you no longer use:

  
  

  ollama rm modelname

  
  

• Pull Fails with Network Error: Don’t panic! Ollama downloads are designed to be resumable. Just re-run the ollama pull command for the model that failed, and it should pick up where it left off.

  
  

FutureFormDigital Insight: Our Recommendation

Keeping Ollama and your models updated is generally a good idea for the best compatibility and access to new AI capabilities. However, if you’re running Ollama as part of a critical workflow or service, we recommend a cautious approach:

1. Test First: If possible, try updating in a non-production environment or on a less critical machine first.
2. Backup (Just in Case): While Docker volumes and Ollama’s design generally keep your models safe during app updates, a quick backup of your model directory (especially if not using Docker volumes, or if you’re extra cautious) before a major application update never hurts. For Docker users, the cp -al method shown in some resources is a good safety net.
3. Update Models During Off-Peak Hours: For large batches of model updates, run the scripts during times when your AI usage is lowest to avoid interruptions.

Most users will find that simply following the OS-specific app update steps and then running ollama pull for their models is smooth sailing. But a little preparedness goes a long way in maintaining those resilient digital workflows!

Frequently Asked Questions (FAQ)

Here are answers to some common questions about updating Ollama:

• Q1: Do I need to uninstall Ollama before updating?
  
  A1: Generally, no. For macOS, Windows, and Linux, the official installers or scripts handle the upgrade process over your existing installation. For Docker, you stop and remove the old container, but the data in the volume is preserved.

• Q2: What happens to my downloaded models when I update the Ollama application?
  
  A2: If you’re using Docker with volumes, or if you’re on macOS, Windows, or Linux where models are stored in their default user directories, your models should be preserved. The Ollama application update process is designed not to delete your model files.

• Q3: Is there a single command to update all my Ollama models at once?
  
  A3: No, not directly. However, you can use command-line scripting on Linux/macOS or PowerShell on Windows to automate pulling all your installed models. (See the “Updating All Models at Once” section above).

• Q4: How do I update Ollama if I installed it using Docker?
  
  A4: You need to pull the latest Docker image (docker pull ollama/ollama:latest) and then recreate your container, making sure to use the same volume mapping (-v your_volume_name:/root/.ollama) as before.

• Q5: My Ollama service won’t start after an update on Linux, what should I do?
  
  A5: Check the service logs using journalctl -u ollama -n 100. Sometimes, old library files can cause issues. Try removing old Ollama library directories (sudo rm -rf /usr/lib/ollama) and re-running the install script.

• Q6: How can I check which Ollama version I’m running?
  
  A6: Use the command ollama --version in your terminal, or query the API with curl http://localhost:11434/api/version.

• Q7: What if a model behaves strangely after updating Ollama?
  
  A7: The interaction templates might have changed. Try re-pulling the specific model using ollama pull <model_name>.

• Q8: How much disk space do Ollama models typically take?
  
  A8: It varies greatly. Smaller models might be a few gigabytes, while larger ones (like 70B parameter models) can easily be 40GB or more. Always check the model’s requirements.

• Q9: Can I revert to an older version of Ollama if the new one has issues?
  
  A9: On Linux, you can use the OLLAMA_VERSION environment variable with the install script (e.g., curl -fsSL https://ollama.com/install.sh | OLLAMA_VERSION=0.3.14 sh). Be aware that versions after 0.4.0 changed the model storage format, so downgrading might require re-pulling models.

• Q10: How do I update Ollama on Windows?
  
  A10: Download the latest installer from ollama.com/download and run it. It will update your existing installation in place and preserve your models.

  
  

Wrapping Up

Keeping your Ollama setup current is a vital part of maintaining a powerful, independent AI workflow. By understanding the difference between application and model updates, and by following the platform-specific steps, you can ensure your AI journey is smooth and uninterrupted.

So, what’s your strategy for managing Ollama updates? Do you prefer to update as soon as a new version drops, or do you wait for a bit of community feedback? Let us know in the comments below – we’d love to hear about your experience and any tips you have!

About The Author