Ollama: 'Error: accepts 1 arg(s), received 2' from pull, show, create or stop

Error: accepts 1 arg(s), received 2 is the Ollama command line counting the words after the command and finding more model names than that command takes. Give pull, show, create or stop one model name, and run the command once per model when you have several.

The fix is to drop the extra word or loop over the names: ollama pull qwen2.5:0.5b, then ollama pull all-minilm. The same message with other numbers (accepts 2 arg(s), received 1) means a word is missing.

The error

PS> ollama pull qwen2.5:0.5b all-minilm
Error: accepts 1 arg(s), received 2
PS> ollama show qwen2.5:0.5b all-minilm
Error: accepts 1 arg(s), received 2
PS> ollama create mymodel -f Modelfile extra
Error: accepts 1 arg(s), received 2
PS> ollama stop qwen2.5:0.5b all-minilm
Error: accepts 1 arg(s), received 2
PS> ollama cp tutor
Error: accepts 2 arg(s), received 1
PS> ollama pull
Error: accepts 1 arg(s), received 0

Why it happens

Each Ollama subcommand declares how many plain arguments it takes. pull, show, create and stop took exactly one in the test, cp took exactly two (source and destination), and the client refused anything else. It did not need the server for that: with OLLAMA_HOST pointed at a port where nothing listened, ollama pull qwen2.5:0.5b all-minilm printed the same message. Flags such as -f Modelfile are not counted, so in the create line the stray word extra is the second argument. Every refused command returned exit code 1.

Listing several models in one ollama pull, the way apt-get install a b c takes several packages, is one way to meet it. Not every command is strict: ollama rm tutor2 tutor3 deleted both models in the same session.

The fix

PS> ollama show qwen2.5:0.5b
  Model
    architecture        qwen2
    parameters          494.03M
...
PS> foreach ($m in 'qwen2.5:0.5b','all-minilm') { ollama pull $m }
...
success
...
success
PS> ollama rm tutor2 tutor3
deleted 'tutor2'
deleted 'tutor3'

One name per call worked for show, pull, create, run and stop, and two names for cp (ollama cp tutor4 tutor5 printed copied 'tutor4' to 'tutor5'). In PowerShell a foreach loop pulls a list, as above. For cp, pass both the source and the new name. If the extra word was meant as a prompt, that belongs to ollama run, which takes the model name and then the prompt text: ollama run tutor5 "Say hi in three words" answered Hi there!.

How it was reproduced

On Windows 11 in Windows PowerShell 5.1 with Ollama 0.40.16, against a separate test server started with OLLAMA_HOST=127.0.0.1:15190 and its own models folder holding qwen2.5:0.5b, all-minilm and two copies named tutor2 and tutor3. Each wrong command was run once and its output and exit code were saved; then the single-name forms, the pull loop and the two-name rm were run against the same server.

Frequently asked

How do I pull several Ollama models at once?

ollama pull takes one model name. Run it once per model, for example in a PowerShell foreach loop or a bash for loop over the names.

What does 'accepts 2 arg(s), received 1' mean in Ollama?

The command needs two plain arguments and got one. ollama cp is the example: it needs the source model and the new name.

Does ollama rm accept more than one model?

In the test with Ollama 0.40.16, ollama rm tutor2 tutor3 deleted both models, unlike pull, show, create and stop, which refused a second name.

More decoded errors in the Fixes category. For a model name that Ollama refuses for its spelling rather than its count, see Ollama: 'invalid model name'.