implementation detail · filed under model architecture
Mixture-of-Experts layers
Transformer layers that use a mixture of experts, including arrangements interleaved with other layer types.
Also called MoE layers.
- sources
- 2
- models
- 2
- labs adopt it
- 2
- strongest
- core
How sources treat it
One count per evidence span, weakest treatment to strongest.
used 1core 1
Documented in
Evidence
2 spans quoted from the sources, strongest treatment first.
interleaved Mamba-2 and MoE layers
coremodel architecturein Nemotron 3 UltraNVIDIA
We employ MoE layers in all Transformer blocks, using SwiGLU activation function with clamping (OpenAI, 2025) at a threshold of 10.
usedunclearin DeepSeek-V4.1-FlashDeepSeek
Filed alongside
Other methods under model architecture :: channel mixer :: mixture of experts.
Mixture of ExpertsSparse expert activationDeepSeekMoEGated DeltaNet MoEMoE with routed and shared expertsAsymmetric input/output activation splitDispatch recomputationFrequency-based partitioning of N-gram embedding slotsHybrid Mixture of ExpertsMegaMoEMoE with 128 routed experts and a shared expertMoE with 256 experts and top-8 routingMoE with 384 routed experts and a shared expertOutput-independent MoE gradient reformulationRouted expert output modulationSigmoid-based MoE router with auxiliary-loss-free load balancing