$2800 rig with 8x Radeon Pro V620 (256 GB VRAM) + custom vLLM fork = Qwen3.8-Flash-Next at 60 to 100 t/s decode and 3000+ t/s prefill

1. 

   消息推送

   $2800 rig with 8x Radeon Pro V620 (256 GB VRAM) + custom vLLM fork = Qwen3.8-Flash-Next at 60 to 100 t/s decode and 3000+ t/s prefill

   02:05 · 2026年10月9日 · 周五

   $2800 rig with 8x Radeon Pro V620 (256 GB VRAM) + custom vLLM fork = Qwen3.8-Flash-Next at 60 to 100 t/s decode and 3000+ t/s prefill
   
   🌐 链接: https://www.reddit.com/r/LocalLLaMA/comments/1x0wnz1/2800_rig_with_8x_radeon_pro_v620_256_gb_vram/
   🏷️ 分组: Reddit｜AI·自托管·VPS
   🕒 时间: 2026-10-09 01:12:46

   Reddit

   

   From the LocalLLaMA community on Reddit: $2800 rig with 8x Radeon Pro V620 (256 GB VRAM) + custom vLLM fork = Qwen3.8-Flash-Next…

   Explore this post and more from the LocalLLaMA community

   02:05 · 2026年10月9日 · 周五