Attempting to use a Gemini Model causes an error.

[Fish-Owl]

I attempted to switch from the base Qwen model to a Gemini model, and it resulted in the following error:
"The operation couldn't be completed. (ProfileBuilderCore.ACPFCLIError error 1.)" See attached.

Selecting "Prepare" and then restarting seemed to fix things.

[johnbean393]

@Fish-Owl

Will look into this. Thanks for the feedback.

[johnbean393]

@Fish-Owl

I'm making a new release version 1.1. This won't fix the your issue here, but it'll give us a better debug message as to what's going wrong.

[jjas0n]

Confirming the issue in 1.1. Here is a screenshot of the error messages

[johnbean393]

@jjas0n

Perfect! This will help me narrow down the problem.

[menesescarlos]

I got the same issue

[Marcelony]

Yep, me too. By the way, will the Gemma model be faster than the Qwen models? 2B is fast, but it's not quite instantaneous. Still a very good tool.

[johnbean393]

@Marcelony @menesescarlos

A release with the fix will probably ship in the text 24 hours.

That being said, no, I think even Gemma 4 E2B will not beat Qwen3.5 2B in inference speed.

Inference wise, this use case is decode dominated, so given the same active parameters and device specs, the maximum number of forward passes should be the same. But it's possible that I haven't considered all the factors.

[johnbean393]

@Fish-Owl @Marcelony @menesescarlos @jjas0n

This fix should have shipped in the new release v1.2. You can download the latest version and try it out.

Credit to @raphaelbarreiros for the PR with the fix!