How LatentMoE Reshapes the MoE Paradigm
青稞AI · wechat · 2026-07-19
Using Kimi K3 and NVIDIA Nemotron3 Super as examples, this article explains why LatentMoE could be the key paradigm for the next generation of MoE models.
The core argument is that traditional MoE bottlenecks aren't just about the number of experts. When routing tokens to experts, large-dimension weights must be read and all-to-all communication is required—costs that scale linearly as the model grows. LatentMoE addresses this by first projecting tokens from their original dimension d into a smaller latent space ℓ, performing routing and expert computations in this lower-dimensional space, and then projecting back. This allows it to:
- Reduce expert weight reads from d×m to ℓ×m
- Proportionally decrease all-to-all communication volume
- Activate more experts and expand expert combination space under the same inference cost
The article details two case studies:
- Nemotron3 Super: Features 512 experts, Top-22 routing, ℓ = 1024, and combines Mamba-2, minimal attention, and MTP for inference acceleration.
- Kimi K3: Features 896 experts, Top-16 routing, and an extremely high sparsity ratio. It introduces QuantileBalancing—using quantiles instead of heuristic hyperparameters for load balancing—to maintain training stability at a 2.8T scale.
The author concludes that while NVIDIA leans towards "efficiency first" and Kimi pushes for "extreme scale," both are essentially compressing expert computation dimensions to trade saved resources for greater expert combinatorial capacity. The article also looks ahead to adaptive compression ratios, dynamic Top-K routing, and incorporating structured sparsity into architectural design.
More from Infra
- Anthropic S-1: training infra edge cut per-task cost 40% and doubled revenue growth — BenBajarin · 2026-10-26
- Rewriting scaling law study: optimal synthetic data share rises from 10% to 32% with compute — heghbalz · 2026-10-26
- US states halt new data centers as 72% of Americans oppose local builds — YvesMulkers · 2026-10-26
- $5M Seed Round Can't Afford One DeepSeek Training Run: GPU-less AI Startup's Three Paths — FinanceYF5 · 2026-10-26
- HBM demand shifts: low-stack inference plans shelved, Rubin to go 12-high only — BenBajarin · 2026-10-26
- CircuitHub unveils The Grid, a factory-scale robot making PCBs in the US twice as fast as Shenzhen — garrytan · 2026-10-26