Event ·Multi-source

DeepSeek V4 Flash 家用部署实测：单机峰值 1875 tok/s

Alec Fong 在单台 GB300 DGX Station 上部署 DeepSeek V4 Flash，测得聚合峰值吞吐 1875 tok/s，单流 300 tok/s，128k 上下文下首 token 延迟 5 秒。推文感叹能在家部署该模型的人相当于 80 年代万元户，令人羡慕。

Current read

有人晒了 DeepSeek V4 Flash 在 DGX Station 上的实测数据，单机每秒能跑 1875 个 token，128k 上下文首字只要 5 秒，你可以拿这个数去衡量本地部署值不值。

9 sources53° AI heatLast updated 2026/8/7 07:12:03

Evidence

9 sources

Primary sourceGeek

2026/8/7 07:12:03View source

Related source 1NVIDIA AI

2026/8/5 19:35:09View source

Related source 2pandaily

2026/8/6 02:18:01View source

Related source 3IT之家

2026/8/6 01:49:50View source

Related source 4AI Will

2026/8/6 06:47:44View source

Related source 5Decoder

2026/8/6 16:33:59View source

Related source 6arXiv: DeepSeek

2026/8/7 05:05:05View source

Related source 7ollama

2026/8/8 06:24:31View source

Related source 8歸藏(guizang.ai)

2026/8/7 17:33:38View source

Conflict check

Dedup data only shows that these sources discuss the same event, not that they agree. There is no structured supporting or refuting evidence yet, so no conflict is inferred automatically.

Open the story