Tagged "dynamic-expert-caching"

• Dynamic Expert Cache in llama.cpp Achieves 27% Faster Inference on Large MoE Models 15 April 2026