Eval request: MaxResDefault001/gemma-4-26B-A4B-opus-reasoning-GGUF (abliterated)

#698
by MaxResDefault001 - opened

Hi, and thanks for keeping the leaderboard going. I know testing time is limited, so no pressure at all.

https://huggingface.co/MaxResDefault001/gemma-4-26B-A4B-opus-reasoning-GGUF

It's a small learning project: a LoRA on google/gemma-4-26B-A4B-it (QAT) trained on angrygiraffe's Opus reasoning traces, then abliterated with p-e-w's Heretic. The repo has two Q4_K_M GGUFs, and the one I'd most like tested is gemma-4-26B-A4B-opus-reasoning-abliterated-Q4_K_M.gguf. It uses the standard gemma-4 template, and works with thinking on or off. The LoRA adapter is in the repo too (adapter/), if applying it to the base is easier than running the GGUF.

Thanks!

Sign up or log in to comment