mradermacher

/

model_requests

Like 204

Hi! Would it be possible to add an i1-IQ4_NL quant for Ateron/Gemma-4-Dark-Thoughts-V2-31B?

I’ve tested the current i1-IQ4_XS, i1-Q4_K_S, and i1-Q4_K_M versions on my RTX 5080 16GB. Q4_K_S is almost as fast as IQ4_XS for me, while Q4_K_M is noticeably slower.

I previously used IQ4_NL on Ateron’s Novelist-Eclipse-31B and had very good results with it, with speed close to IQ4_XS. I’d be curious to compare IQ4_NL with Q4_K_S on Dark Thoughts V2 as well.

Thanks for providing all these GGUF quants!