Event ·Multi-source

Cursor开源MoK：MoE训练内核速度提升2.37倍

Cursor宣布开源Mixture-of-Kittens (MoK)，这是一个面向NVL72的MoE训练megakernel。MoK将专家混合的所有通信与计算融合进单个确定性内核，相比最强公开基线提速最高2.37倍。该内核专为NVL72集群设计，用于提升大规模MoE训练效率。

Current read

Cursor开源了MoK内核，把MoE训练的通信和计算合成一个内核，跑得比公开基线快2.37倍，搞NVL72训练的可以看看。

3 sources72° AI heatLast updated 2026/8/4 19:18:37

Evidence

3 sources

Primary sourceeric zakariasson

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

Related source 1Cursor

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

Related source 2marktechpost

2026/8/4 18:38:41View 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