support for ollama for the model: unsloth/Qwen3-4B-Thinking-2507-GGUF

[Aryabhattacharjee]

I see the benchmark. Its a great model.
But in huggingface repo, in use this model dropdown, i don't find ollama. Also cant convert this model to ollama also.
Please provide support for ollama for this model.

link of repo:
https://huggingface.co/unsloth/Qwen3-4B-Thinking-2507-GGUF

[shimmyshimmer]

Someone said that it already works in Ollama for the Thinking one. We did add it for the instruct model already

[slepkaviba]

Hey, in Instruct template was added and it was fixed, will same changes apply to Thinking?

[slepkaviba]

@mmathew23 - I did check that there is Ollama in the dropdown, but templates are missing, so Thinking haluciates same way as Instruct did before update.

[mmathew23]

Hi @slepkaviba we just added template and param files. Please try again.

[slepkaviba]

Hello @mmathew23 - thank you for quick reply, it seems to work "mostly", but sometimes it "stucks" for a while before it starts to think...

Not sure if that is template, or my local setup.

[rolandtannous]

In addition to what mathew already shared, we also recently pushed a major fix for llamacpp integration.

1- Please make sure to update your local install of unsloth

pip install --force-reinstall --no-deps unsloth-zoo unsloth

If you're using a notebook on kaggle or colab, it should pull the latest version automatically.

If the model is supported by Ollama, it will now also automatically create a Modelfile for you that you can use to create your model in ollama using:

ollama create model_name -f ./Modelfile

Will be closing this for now as the issue has been resolved. Please make sure to follow the instructions given in my last response.

Meanwhile , please feel free to respond if you're still having any issues.

Thank you