Model Management
Browsing Local Models
The Models → Local tab lists all models you have pulled. For each model you can see:
- Parameter count, quantization, and file size
- GPU/VRAM requirements (detected from your hardware via
detect_hardware) - Custom tags and favorite status
Click a model to open its detail panel, where you can view the full Ollama model card, available tags, and edit per-model generation defaults.
Pulling Models
- Go to Models → Library and search for a model name.
- Click a model to see available tags (e.g.
7b,13b,instruct,q4_K_M). - Click Pull to start downloading. Progress is shown in real time.
- The model appears in the Local tab once complete.
You can also pull directly by name: open a conversation, type the model name into the model selector, and Alpaka Desktop will prompt you to pull it.
Deleting Models
In Models → Local, click the three-dot menu on any model → Delete. The model is removed from disk via ollama rm.
Tags and Favorites
- Favorite a model (⭐) to float it to the top of the model selector.
- Tags are custom text labels you apply to models (e.g.
coding,fast,vision). - Filter the local model list and the chat model selector by tag.
Custom Model Creation (Modelfile)
Alpaka Desktop includes a full Modelfile editor (CodeMirror) for creating custom Ollama models:
- Go to Models → Create.
- Write your Modelfile (see the Ollama Modelfile reference).
- Click Create. Progress streams in real time; you can cancel mid-stream.
- A desktop notification confirms when the model is ready.
Storage Path
By default, Ollama stores model blobs in ~/.ollama/models. To change this:
- Go to Settings → Engine → Model Storage Path.
- Enter the new absolute path.
- Click Save. Alpaka Desktop writes a systemd service override (
OLLAMA_MODELS) and restarts the Ollama service automatically.
WARNING
Changing the storage path does not move existing models. You must manually move the ~/.ollama/models/ directory contents to the new path before saving, or re-pull models after the change.