I’m going to have to test this out later since it looks like what I’ve been trying to do with other LLMs with mixed results.

https://byteshape.com/blogs/Qwen3-30B-A3B-Instruct-2507