Event ·Multi-source

Cursor 开源 MoK：NVL72 MoE 训练内核提速 2.37 倍

Cursor AI 开源了 Mixture-of-Kittens (MoK) 训练内核。该内核面向 NVIDIA NVL72 平台，将全部 MoE 通信与计算融合为单个确定性内核。相比最强公开基线，MoK 的训练速度最高可提升 2.37 倍。开发者现在可以获得完整代码并用于自己的训练任务。

Current read

Cursor 把 NVL72 上的 MoE 训练内核开出来了，比公开方案快 2.37 倍，想省训练成本直接用。

11 sources72° AI heatLast updated 2026/8/4 16:00:26

Evidence

11 sources

Primary sourceCursor

2026/8/4 16:00:26View source

Related source 1eric zakariasson

2026/8/4 19:18:37View source

Related source 2LangChain

2026/8/4 13:04:48View source

Related source 3marktechpost

2026/8/4 18:38:41View source

Related source 4NVIDIA AI

2026/8/3 16:07:28View source

Related source 5Decoder

2026/8/4 12:23:43View source

Related source 6arXiv cs.LG

2026/8/4 14:01:00View source

Related source 7Hugging Face

2026/8/4 21:53:10View source

Related source 8IT之家

2026/8/5 01:06:17View source

Related source 9elvis

2026/8/5 16:27:02View source

Related source 10Thomas Wolf

2026/8/5 19:25:10View 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