Loading market data...
ai

Cursor Open-Sources Mixture-of-Kittens (MoK): A Deterministic MoE Training Megakernel for GB300 NVL72 Racks

MarkTechPost
Read Full Article at MarkTechPost
Share:PostShare
Cursor Open-Sources Mixture-of-Kittens (MoK): A Deterministic MoE Training Megakernel for GB300 NVL72 Racks
Ad Slot — In-Article (728x90)

Cursor Research has open-sourced Mixture-of-Kittens (MoK), the MoE training megakernel behind its Composer models. MoK fuses all mixture-of-experts communication and computation into a single deterministic kernel, and runs up to 2. 37x faster than the strongest public baseline on GB300 NVL72 racks.

Cursor Research has open-sourced Mixture-of-Kittens (MoK), the MoE training megakernel behind its Composer models. MoK fuses all mixture-of-experts communication and computation into a single deterministic kernel, and runs up to 2. 37x faster than the strongest public baseline on GB300 NVL72 racks.

This is a summary. For the full story, read the original article at MarkTechPost.

Original source: MarkTechPost

Ad Slot — Below Article (300x250)