deepseek-v4.1POST
4 块 RTX 6000 Pro 本地部署 V4.1 Flash 1017，对比 GLM 5.3 Flash
作者用 YourLocalAILab 的 vLLM fork 在 4 块 RTX 6000 Pro 上跑 V4.1 Flash 1017，并与现有 GLM 5.3 Flash 环境对比，结论是新 DeepSeek 模型成为其本地首选。
Austere
@commanderdata8510 天前收录原推发布于 2026.10.18
Austere @commanderdata85原推发布于 2026.10.18
Sharing the results of my own testing using @YourLocalAILab’s fork of vLLM with the new DeepSeek V4.1 Flash 1017 model, and comparing to my current setup with GLM 5.3 Flash. In short, the new DeepSeek model is now my local model of choice on a server with 4x RTX 6000 Pro GPUs. https://github.com/commdata2338/deepseek-v4.1-flash-1017-4x-rtx-pro-6000-benchmark
x.com/commanderdata85/status/2111679772711321680采集于 2026.10.18 16:44