Error: couldn't find model "nosuchmodel" to stop comes from ollama stop when the name you gave is not a model on that server. Run ollama ps, copy the name from its NAME column exactly, and stop that.

A model that exists but is not loaded is a different case: ollama stop printed nothing and returned exit code 0 for it. So the error points at the name, not at whether the model was running.

The error

PS> ollama stop nosuchmodel
Error: couldn't find model "nosuchmodel" to stop

The command returned exit code 1.

Why it happens

ollama stop takes one model name. In the test, the name nosuchmodel matched no model on the server the client was pointed at, and the message repeated that name in quotes. Compare the quoted name with the output of ollama ps or ollama list character by character, tag included.

The other case is quieter. qwen2.5:0.5b was on the server but not loaded at the time, and ollama stop qwen2.5:0.5b produced no output and exit code 0. A script that checks the exit code therefore cannot tell "stopped" from "was not running", and needs ollama ps before and after if that matters.

The fix

PS> ollama ps
NAME          ID              SIZE      PROCESSOR    CONTEXT    RUNNER      UNTIL
qwen3:0.6b    7df6b6e09427    2.4 GB    100% GPU     4096       llamacpp    4 minutes from now
PS> ollama stop qwen3:0.6b
PS> ollama ps
NAME    ID    SIZE    PROCESSOR    CONTEXT    RUNNER    UNTIL

ollama ps lists the loaded models with their full names, tag included. After ollama stop qwen3:0.6b (exit code 0, no output) the list was empty, which is the confirmation that the model left memory. If the list is empty before you start, there is nothing to stop.

Point the client at the right server as well. With qwen3:0.6b loaded on the test server at 127.0.0.1:15190, setting OLLAMA_HOST=127.0.0.1:15191 (a second test server with the same models folder) made ollama ps list nothing, and ollama stop qwen3:0.6b there printed nothing and exited 0 while the model stayed loaded on 15190.

How it was reproduced

On Windows 11 in Windows PowerShell 5.1 with Ollama 0.40.16, against a separate test server on 127.0.0.1:15190 with its own models folder. qwen3:0.6b was loaded (ollama ps listed it); qwen2.5:0.5b existed but was not loaded. The client ran ollama stop with the unloaded model, with a name that does not exist, then ollama ps, ollama stop qwen3:0.6b and ollama ps again.

Frequently asked

Why does ollama stop say couldn't find model to stop?
The name given matched no model on the server the client talked to. Copy the name from the NAME column of ollama ps and try again.
What happens if I ollama stop a model that is not running?
In the test with Ollama 0.40.16, stopping a model that existed but was not loaded printed nothing and returned exit code 0.
How do I check that ollama stop worked?
Run ollama ps afterwards. In the test the list was empty after ollama stop qwen3:0.6b.

More decoded errors in the Fixes category. For ollama stop with two names at once, see Ollama: 'accepts 1 arg(s), received 2'.