Raon-1-mini-preview (28.8B/4.3B MoE) on hipfire, 7900 XT: runs natively, but ~200× slower prefill / ~10× slower decode than llama.cpp

1. 

   消息推送

   Raon-1-mini-preview (28.8B/4.3B MoE) on hipfire, 7900 XT: runs natively, but ~200× slower prefill / ~10× slower decode than llama.cpp

   2小时前

   Raon-1-mini-preview (28.8B/4.3B MoE) on hipfire, 7900 XT: runs natively, but ~200× slower prefill / ~10× slower decode than llama.cpp
   
   🌐 链接: https://www.reddit.com/r/LocalLLaMA/comments/1xitxzt/raon1minipreview_288b43b_moe_on_hipfire_7900_xt/
   🏷️ 分组: Reddit｜AI·自托管·VPS
   🕒 时间: 2026-10-30 03:40:29

   Reddit

   

   From the LocalLLaMA community on Reddit: Raon-1-mini-preview (28.8B/4.3B MoE) on hipfire, 7900 XT: runs natively, but ~200× slower…

   Explore this post and more from the LocalLLaMA community

   2小时前