Ollama Update Makes Your AI Models Faster and Safer

Ollama, a free tool that lets you run AI models like Llama and Mistral directly on your laptop, just got a small but important update. The new version, v0.40.2, automatically upgrades any models you downloaded with older versions the first time you run them. This upgrade makes the models run faster and work better with the underlying technology, llama.cpp. You don't have to do anything—it happens in the background.

But there's a trade-off: to keep your options open, Ollama keeps a backup of the original model files. That means your disk space will temporarily be used up by both the old and new versions. If you're tight on storage, this could be a problem. The update includes instructions to manually delete these backups using a command-line tool called jq. However, if you later decide to downgrade to an older version of Ollama, you'll need to re-download those models.

For most users, this update is a net positive. Faster models mean quicker responses and smoother performance, especially on older computers. The automatic upgrade saves you the hassle of manually updating each model. Just be aware that your hard drive might get a bit fuller until you clean up the backups. A future release will handle that cleanup automatically, but for now, it's a manual step for those who want to reclaim space.

Overall, this is a maintenance update that improves the experience for anyone running AI locally. It's not a flashy new feature, but it makes the tool more reliable and efficient. If you use Ollama, you'll likely notice better performance without lifting a finger.