This link [1] features some good insight on how to adapt your usage to smaller models which require more explicit or deliberate prompting. I have been using Gemma 4 31B a lot and have found it very competent. It can be a bit unstable and start spiraling or end up in infinite loops that you need to reset, but for the most part it's been really good.

[1]: https://itayinbarr.substack.com/p/honey-i-shrunk-the-coding-...