AI Usage Tracker

Why Is Ollama ps Empty When Ollama List Shows My Model?

An empty Ollama ps list can be normal even when your model is installed. Check the model’s loaded state, idle timeout and server address on your Mac.

You can see your model in ollama list and get only column headers from ollama ps. Before pulling it again or starting another Ollama process, check what happens when you send a prompt to the server your terminal is querying.

On this page

1. Check the same model before and after a prompt

Run ollama list first. In the next commands, replace gemma4:e4b with the full tag of a local text model already in your list. Use a model you normally run on this Mac.

ollama list
ollama run gemma4:e4b "Reply with OK." && ollama ps

A successful prompt followed by a row for the model tells you that Ollama loaded it and still has it in memory. The row can stay visible while the model is idle. Ollama documents ollama ls as another form of the installed-model command. Ollama CLI reference

Example readings for an installed model
Before use
list shows the tag
ps has no row
After a prompt
list shows the tag
ps shows the tag
After idle unload
list shows the tag
ps has no row

2. Check the idle timeout

If the model appears and then disappears, look at the UNTIL column while it’s loaded. Ollama defaults to five minutes in memory after use. A server can set another duration with OLLAMA_KEEP_ALIVE, and a client can override it for a request. If the client sends keep_alive: 0, Ollama unloads the model as soon as the response finishes. Ollama FAQ on keeping models loaded

3. Check which server your client uses

If a chat app answers while ollama ps stays empty, compare the app’s Ollama server address with OLLAMA_HOST in your terminal. The default local address is http://127.0.0.1:11434. An app using a different Mac, port or container can get a response while your local server has no model loaded. Ollama network settings

To isolate the endpoint and timeout, use the same Ollama server for both requests. Set base to the server’s root URL, with no /v1 or /api suffix, and use the same installed tag as before. The test asks Ollama to hold the model for two minutes after the prompt.

base=http://127.0.0.1:11434
curl -fsS "$base/api/generate" \
  -H 'Content-Type: application/json' \
  -d '{"model":"gemma4:e4b","prompt":"Reply with OK.","stream":false,"keep_alive":"2m"}' &&
  curl -fsS "$base/api/ps"

The model name in the second response confirms it’s loaded on that server. A connection error means you need to check the address or whether Ollama is running there. If the test works but your app still leaves the list empty, check the app’s address and the keep-alive value it sends. Generate API · List running models API