List of all available models

[vt-alt]

Please provide list of (or API to list) all models available on https://ollama.com/library with tags.
This would be useful for users to get them from cli without a browser.
Also, this would be useful for shell completions.

[vt-alt]

I see that request is reappearing. Models list could be just a file without internet access packed on every release into docs.

[rick-github]

Until the API is implemented, #7239 (comment) has some code that scrapes the library.

[kth8]

I've been using these 2 one liners in a justfile to quickly get models and tags

set positional-arguments := true

default:
  just --list

#list Ollama models
@models:
    curl -s https://ollama.com/library | grep -oP 'href="/library/\K[^"]+'

#list model tags
@tags model:
    curl -s https://ollama.com/library/$1/tags | grep -o "$1:[^\" ]*q[^\" ]*" | grep -E -v 'text|base|fp|q[45]_[01]'

$ just models
qwen2.5-coder
llama3.2-vision
llama3.2
llama3.1
llama3
mistral
...

$ just tags llama3.2
llama3.2:1b-instruct-q2_K
llama3.2:1b-instruct-q3_K_L
llama3.2:1b-instruct-q3_K_M
llama3.2:1b-instruct-q3_K_S
llama3.2:1b-instruct-q4_K_M
llama3.2:1b-instruct-q4_K_S
llama3.2:1b-instruct-q5_K_M
llama3.2:1b-instruct-q5_K_S
llama3.2:1b-instruct-q6_K
llama3.2:1b-instruct-q8_0
llama3.2:3b-instruct-q2_K
llama3.2:3b-instruct-q3_K_L
llama3.2:3b-instruct-q3_K_M
llama3.2:3b-instruct-q3_K_S
llama3.2:3b-instruct-q4_K_M
llama3.2:3b-instruct-q4_K_S
llama3.2:3b-instruct-q5_K_M
llama3.2:3b-instruct-q5_K_S
llama3.2:3b-instruct-q6_K
llama3.2:3b-instruct-q8_0

[vt-alt]

I've been using these 2 one liners in a justfile to quickly get models and tags

Thanks. Wow but it's so slow and definitely not suitable for bash completions as run-time http requests. But of course, list could be prepared in advance.

[kth8]

Scraping the whole library takes 0.8s and tags take 0.3s from my benchmark

$ hyperfine "just models"
Benchmark 1: just models
  Time (mean ± σ):     813.1 ms ±  79.3 ms    [User: 51.7 ms, System: 14.7 ms]
  Range (min … max):   634.0 ms … 887.5 ms    10 runs

$ hyperfine "just tags llama3.2"
Benchmark 1: just tags llama3.2
  Time (mean ± σ):     285.3 ms ±  87.1 ms    [User: 50.9 ms, System: 14.9 ms]
  Range (min … max):   199.6 ms … 530.8 ms    11 runs

I'm not trying to use this for autocomplete, mainly used it for automated benchmarking

Ollama-MMLU-Pro $ just tags llama3.2 > models.txt
Ollama-MMLU-Pro $ while read line; do ollama pull $line; pipenv run python run_openai.py -u http://127.0.0.1:11434/v1 -p 3 --category other -m $line; done < models.txt

[vt-alt]

Yes now it's much faster, perhaps some glitch at the time I tested it. Thanks

[dhiltgen]

Looks like a dup of #286