The sparse Mixture-of-Experts (MoE) architecture is increasingly favored for scaling Large Language Models (LLMs) efficiently, but it depends on heterogeneous compute and memory resources. These factors jointly affect system Cost, Accuracy, and Performance (CAP), making trade-offs inevitable. Existing benchmarks often fail to capture these trade-offs accurately, complicating practical deployment decisions. To address this, we introduce MoE-CAP, a benchmark specifically designed for MoE systems. Our analysis reveals that achieving an optimal balance across CAP is difficult with current hardware; MoE systems typically optimize two of the three dimensions at the expense of the third-a dynamic we term the MoE-CAP trade-off. To visualize this, we propose the CAP Radar Diagram. We further introduce sparsity-aware performance metrics-Sparse Memory Bandwidth Utilization (S-MBU) and Sparse Model FLOPS Utilization (S-MFU)—to enable accurate performance benchmarking of MoE systems across diverse hardware platforms and deployment scenarios.
MoE-CAP introduces a comprehensive benchmarking framework for sparse Mixture-of-Experts systems with three main contributions:
A novel visualization tool that displays the trade-offs between Cost, Accuracy, and Performance across different MoE systems and hardware configurations.
The benchmark addresses critical gaps in existing MoE evaluation by:
MoE-CAP enables practitioners to:
This work provides essential tools for the growing ecosystem of sparse MoE systems, helping bridge the gap between theoretical model capabilities and practical deployment realities.