Ollama 0.40.2 upgrades downloaded models for llama.cpp

Ollama v0.40.2 was published on October 8, 2026 as a release of the ollama/ollama repository on GitHub. According to the release notes, models downloaded with earlier versions are upgraded in the background the first time they run, for better performance and compatibility on llama.cpp.

Release facts

What changed

• Background model upgrade. Models downloaded with earlier Ollama versions are upgraded the first time you run them, for better performance and compatibility when running on llama.cpp.
• Backups for safe downgrades. Ollama keeps the original copy as a backup, so upgraded models are temporarily kept on disk. The release notes say a future release will remove these backups automatically.
• Manual backup cleanup. The notes include a shell snippet that requires jq and removes the backed up models now. It only deletes the backups.
• No duplicate list entries. ollama list no longer shows duplicate entries for upgraded models.
• Claude launch context. ollama launch claude uses the model's full context length.

Questions

Do I need to re-pull models after upgrading?

The release notes describe an automatic background upgrade on first run, so no manual step is mentioned for that. Re-pulling is only needed if you delete the backups and later downgrade to a version older than 0.40.

Does the upgrade use extra disk space?

Yes, temporarily. The release notes say the original copy is kept as a backup and that a future release will remove these backups automatically.

Is there a way to remove the backups now?

The release notes provide a command using curl, jq and ollama rm for this purpose.

See the Ollama hub, the coding agents category, and a related launch, Cline SDK v0.0.93.