DeepSeek V4 Flash 0731 IQ2_M benchmark for Dual 3060 and 96GB RAM ≈ 3.5 tok/s.

Thanks to the community help I finally launched this llm. LM Studio refused to load weight onto second GPU but Unsloth Studio did so everything was done in there. Not a proper benchmark (used PC in parallel as well) but it gives an idea of ​​the performance from dual 3060 with RAM offloading.

中文全文翻译正在补齐，可先查看下方英文原文。

Thanks to the community help I finally launched this llm. LM Studio refused to load weight onto second GPU but Unsloth Studio did so everything was done in there. Not a proper benchmark (used PC in parallel as well) but it gives an idea of ​​the performance from dual 3060 with RAM offloading.