Cursor Releases Mixture-of-Kittens Megakernel for MoE, Claims Nearly 2x TFLOP/s

CapnHat · reddit · 2026-08-05

Cursor officially released Mixture-of-Kittens, a new megakernel designed specifically for training Mixture-of-Experts (MoE) models. The company claims the technology can nearly double TFLOP/s performance, maximizing hardware utilization for MoE architectures.

The project is open-source on GitHub, allowing developers to test and squeeze more performance out of their preferred MoE models.

Related event: Cursor Open-Sources MoK Megakernel, Nearly Doubling NVL72 MoE Performance(5 posts)→

Original post →

More from Infra

Infra channel →