Outmake Beta Download Beta
Guides

Use your LM Studio and Ollama models in Outmake

Model files are big, and every app that runs them wants its own download. Outmake uses the file you already have. It takes no extra space on your Mac, and the other app keeps working.

How it works

Press Download on a chat, picture or video model in Outmake. Before it downloads anything, Outmake looks for the same file in the folders other apps keep models in. If it finds one, it checks the file against the checksum in Outmake’s catalog, then uses it. The download is skipped.

Why it takes no extra space

Outmake makes a clone of the file, which is how a Mac’s usual disk format (APFS) copies a file without storing its contents twice. Both apps see their own file. Delete the model in one app and the other still has it; the space comes back when both are gone.

Where it looks

  • LM Studio: ~/.lmstudio/models and ~/.cache/lm-studio/models
  • Ollama: ~/.ollama/models/blobs
  • The Hugging Face cache: ~/.cache/huggingface/hub
  • llama.cpp’s own cache: ~/Library/Caches/llama.cpp
  • ~/.local/share/models, where other llama.cpp tools keep theirs

Only these folders. Outmake doesn’t look in Documents, Desktop or Downloads.

When it downloads anyway

  • The file isn’t the same one. A different size of the same model (Q4 where Outmake wants Q6) is a different file, and so is one with another name or size.
  • The checksum doesn’t match. Any app can write to those folders, so a file is never trusted by its name.
  • The file is on another disk. A copy across disks would take the space again.

Keep running them in LM Studio or Ollama

If you’d rather leave a model running in LM Studio or Ollama and chat with it from Outmake, OpenCode in Outmake can talk to either one’s server on this Mac.

Going the other way, every chat model’s page in Outmake’s model list has a link that opens it in LM Studio, Jan or Msty, and the Ollama and llama.cpp commands for it.

More about models on this Mac