Lawyer with too many GPUs, part 4: DeepSeek-V4.1-Flash on eight V100s from 2018. 200+ tok/s on code, ~110 on legal writing, 2,200 tok/s prefill at 128K, 524K context, every weight in VRAM

1. 

   消息推送

   Lawyer with too many GPUs, part 4: DeepSeek-V4.1-Flash on eight V100s from 2018. 200+ tok/s on code, ~110 on legal writing, 2,200 tok/s prefill at 128K, 524K context, every weight in VRAM

   4小时前

   Lawyer with too many GPUs, part 4: DeepSeek-V4.1-Flash on eight V100s from 2018. 200+ tok/s on code, ~110 on legal writing, 2,200 tok/s prefill at 128K, 524K context, every weight in VRAM
   
   🌐 链接: https://www.reddit.com/r/LocalLLaMA/comments/1xoubyo/lawyer_with_too_many_gpus_part_4_deepseekv41flash/
   🏷️ 分组: Reddit｜AI·自托管·VPS
   🕒 时间: 2026-11-06 05:07:54

   Reddit

   

   From the LocalLLaMA community on Reddit

   Explore this post and more from the LocalLLaMA community

   4小时前