[fluffyspork]

Gemma 4 31b is better for coding than Gemini in my limited testing on a small C project single source file project, less than 1000 lines. Setting temperature to 0 gives better results for me. It seems like Gemini ignores the system prompt more and the default reasoning output seems more incoherent.

[_fat_santa]

Their open weight on device models are really impressive. Partly because I think they are the only ones out of all the frontier labs even working on local models.

[onlyrealcuzzo]

> Gemma 4 31b is better for coding than Gemini

Is there a fine-tuned Gemma coding model? I'd assume that would perform quite well.

[fluffyspork]

I don't know, but for me setting temperature to 0 has a noticeable impact. I don't think you need a fine tune.