Updating and Managing Ollama Models

What this article covers

• How to bring models up to date.
• How to use tags and versions.
• How to free up disk space.
• How to copy and delete models.
• Best practices for clean model management.

Introduction: Updating and managing Ollama models

Ollama models receive regular updates. New versions bring bug fixes, improved quantization, or extended context length. Without careful tracking, you’ll quickly accumulate multiple versions of the same model on disk and lose sight of what you have. This wastes storage and creates confusion when an outdated model gets used by mistake.

This article explains how to efficiently update, manage, and clean up Ollama models.

Key concepts

• Pull: Download a new model.
• Tag: Version identifier for a model.
• Manifest: Model metadata.
• Blob: Large model file.
• Garbage Collection: Cleanup of unused data.
• Alias: Alternative name for a model.
• Copy: Duplicate a model under a new name.
• Remove: Delete a model.

Listing models

ollama list

This command displays all available models, their tags, and sizes.

Updating models

Download a new version:

ollama pull llama3.1:latest

Ollama downloads only the changed parts, saving time and bandwidth.

Checking for old versions

After a pull, old blobs may still exist. You can identify them with the prune function or manually:

ollama ls

Freeing up space

ollama rm llama3.1:old-version

Delete multiple models at once:

ollama rm model1 model2

Warning: deleted models must be downloaded again.

Copying models

Save a model under a new name:

ollama cp llama3.1 my-llama

This is useful when you want to create variations based on an existing model.

Understanding tags

Tag	Meaning
latest	Most recent official release.
8b, 13b, 70b	Model size.
q4_0, q4_K_M	Quantization level.
v1.2.3	Version number.

Use tags to maintain multiple versions in parallel:

ollama pull qwen2.5:14b
ollama pull qwen2.5:7b
ollama pull qwen2.5-coder:14b

Automatic updates

Ollama has no built-in auto-updater for models. A cron job can run pull on a schedule:

0 4 * * 1 ollama pull llama3.1:latest >> /var/log/ollama-update.log 2>&1

Proceed with caution: new models may behave differently. Reserve automatic updates for latest in test environments only.

Updating models from a Modelfile

If a custom model depends on a base model:

ollama pull llama3.1
ollama create my-assistant -f Modelfile

Version control for Modelfiles

Store Modelfiles in Git or a version-controlled directory:

modelfiles/
  llama3.1-rag-v1.md
  llama3.1-coding-v2.md

Checking disk usage

du -sh ~/.ollama
ollama list

Removing unused models quickly frees up many gigabytes.

Tips

• Don’t use latest in production. Use fixed tags like version numbers instead.
• Clean up regularly.
• Back up important models before changes.
• Version control your Modelfiles.
• Test updates in a test environment first.
• Monitor disk usage.

Common pitfalls

• Wrong tag: Model is older than expected.
• Using only latest: No reproducibility.
• Not cleaning up: Disk fills up quickly.
• Updating without backup: Custom models may be lost.
• Deleting a needed model: Re-downloading takes time.
• Confusing multiple versions: Performance varies between versions.

Further reading

• BotServ.de Ollama commands
• BotServ.de Ollama model lifecycle
• BotServ.de Ollama backup
• BotServ.de Ollama quantization in practice

FAQ: Updating Ollama models

How do I check if an update is available? Run ollama pull and it will tell you if new data needs to be downloaded.

Do I need to delete old models? No, but it saves disk space.

Can I update models automatically? Only through external cron jobs. Ollama itself has no built-in auto-updater.

What do the tags mean? They identify model size, version, or quantization level.

Should I always use latest? Use it for testing, but pin specific tags for reproducible applications.

Sources and further reading

• Ollama Model Library: https://ollama.com/library
• Ollama CLI: https://github.com/ollama/ollama

Summary: Updating and managing Ollama models

Model management in Ollama becomes straightforward once you understand tags and commands. Regular pull updates models, rm cleans up space, and cp lets you create variations. For reproducible workflows, use fixed tags and version-controlled Modelfiles. By tracking disk usage and model versions, you avoid unnecessary downloads and maintain consistency across your work.