Model Architecture & Training MethodsGoogle (Lepikhin et al.)Google

GShard

GShard

Why it matters

Scaled MoE up to a 600B-scale system and established expert parallelism and automatic sharding

Source — Primary sources first; secondary reports are labelled “Press”

Paperarxiv.orghttps://arxiv.org/abs/2006.16668

Added in Week 35, 2026 · See that week

Nearby on the same axis

Switch TransformerPaperFast Transformer DecodingPaperRoPE / RoFormerPaperALiBiPaper