Start or stop a model

Start or stop a model

LM Nexus can start, monitor, restart, and stop a local llama-server for a Model Manager profile.

Start a profile

1. Open Model Manager.
2. Select Profiles.
3. Find the profile you want to run.
4. Use the circular action with the tooltip Start model.

While the process starts, the profile enters a starting/loading state.

Expected result

When startup succeeds:

• the profile status becomes ready/running;
• the Model Manager running count increases;
• the profile exposes its assigned runtime port;
• the running profile name becomes a link to the managed server address.

The profile tooltip also shows runtime details such as context, GPU layers, threads, batch size, and assigned port.

Restart a running profile

Use the restart action when you intentionally want LM Nexus to stop and recreate the managed process with the profile’s current configuration.

If you changed runtime settings that require a new process, restart is normally more appropriate than expecting an already-running process to change in place.

Stop a profile

Use the circular action with the tooltip Stop model.

Wait for Model Manager to confirm that the managed process has stopped before starting another conflicting process or reusing the same resources.

If startup fails

Check:

• the GGUF still exists;
• the selected profile configuration is valid;
• the configured/managed llama-server binary is available;
• the assigned port is not occupied;
• available system/VRAM resources are sufficient;
• Model Manager/runtime logs contain the actual startup error.

If the UI shows a VRAM warning, treat it as a useful estimate rather than a guarantee that the model will or will not start.

See Troubleshooting for a broader checklist.

Which processes LM Nexus controls

LM Nexus should stop only processes it owns.

If you configure an external or self-hosted model server as a provider, LM Nexus does not take over starting or stopping that server.

Next

Use the local API.