Cursor open-sources Mixture-of-Kittens MoE training megakernel for NVL72
Cursor is open-sourcing Mixture-of-Kittens (MoK), a fused Mixture-of-Experts training megakernel for NVIDIA NVL72 systems that combines MoE communication and compute into one deterministic kernel. Cursor reports up to 2.37x versus strong public baselines and a 1.41x end-to-end training throughput gain over its prior DeepEP-based stack in production.
LLMgram editorial selection and synthesis · @llmgram. LLMgram is not the original publisher of this information.
Continue on LLMgram: Open in AI Signal →
Original reporting: We're open-sourcing Mixture-of-Kittens (MoK), our MoE training megakernel for NVL72s. It fuses all Mixture-of-Experts communication and computation into a single, fully determinis…