Models

The Models page lists every model from your backend, merged with live running state and peer info. You can load and unload models, and open a per-model detail view with stats, residency history, recent requests, a config snippet, and a per-key breakdown.

Model list

The table is built from the backend model list and live running-model state:

• Local models can be loaded or unloaded when the backend supports lifecycle actions.
• Peer models show as remote capacity and are not managed by the local load/unload buttons.
• Refresh re-reads model state from the backend.
• Unload all unloads every currently running local model.

If no models appear, check the llama-swap config.yaml that llama-dash points at with INFERENCE_CONFIG_FILE.

Model detail

Open a model to inspect:

• State, kind, context length, TTL, port, and aliases parsed from the model config snippet.
• Load/unload action for local models.
• Recent residency events.
• Recent requests for that model.
• Request stats and per-key usage breakdowns.
• The command/config snippet that declares the model.

The detail sidecar also links to the Playground and the config editor for quick testing and edits.