Event ·Multi-source

Hermes Agent自动部署DeepSeek v4 Flash，双DGX Sparks单流82 tok/s

MiaAI实验室用Hermes agent根据一条X帖子，在一夜之间自动完成了DeepSeek v4 Flash GA (0731)在2块NVIDIA DGX Sparks上的部署。部署使用官方FP8权重，单流推理速度达82 tok/s，峰值约95 tok/s，三会话并发时达到135 tok/s。该方案已发布在GitHub上，并感谢了0xSero提交的PR，后续还会改进部署配方。

Current read

Hermes agent看了条帖子就自动部署好DeepSeek v4 Flash，双DGX Sparks跑出82 tok/s，省人力还快。

11 sources61° AI heatLast updated 2026/8/1 16:05:59

Evidence

11 sources

Primary sourceandrew chen

2026/8/1 16:05:59View source

Related source 1IT之家

2026/7/31 06:02:05View source

Related source 2深度求索 DeepSeek

2026/7/31 06:56:41View source

Related source 3歸藏(guizang.ai)

2026/7/31 07:14:29View source

Related source 4Geek

2026/7/31 07:24:19View source

Related source 5向阳乔木

2026/7/31 09:00:03View source

Related source 6AI Will

2026/7/31 10:42:34View source

Related source 7lmarena.ai

2026/7/31 15:34:05View source

Related source 8elvis

2026/7/31 15:36:52View source

Related source 9小互

2026/8/1 02:46:34View source

Related source 10orange.ai

2026/7/31 09:25:05View 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