Why Llama 4 Maverick (400B MoE) pressures system RAM
Llama 4 Maverick (400B MoE) is Mixture-of-Experts: inference activates 17B active/token, but VRAM/RAM must usually hold the full ~400B expert set for fast routing. At Q4 the weight slab is ~225GB before KV (~0.04GB at 8K) and ~8GB OS/runtime overhead — totaling ~233GB raw, rounded to a 256GB kit. Stretching toward the full 1M-token window multiplies KV far faster than weights; that is the usual “I bought enough RAM for the model but still OOM” failure on Meta Llama MoE pages.
What RAM kit to buy
Shop 256GB-class capacity for Llama 4 Maverick (400B MoE): workstation DDR5 RDIMM/LRDIMM or multi-kit desktop builds, not a single gamer 2×16GB stick. Use our 128GB+ price hubs and RAM Finder; confirm ECC needs for your board. GPU path: Apple Mac Studio (192GB Unified Memory) or Institutional Node (8x H100 / A100) (237GB VRAM class) if you want weights on-device instead of system-RAM offload.
Workload notes
Meta Llama-family models like Llama 4 Maverick (400B MoE) have broad llama.cpp/Ollama support — prioritize stable JEDEC/EXPO kits over unproven XMP outliers for multi-hour serves. At 400B total parameters this is frontier-scale — expect multi-GPU or heavy CPU offload even in Q4; the 256GB kit is a host-memory floor, not a promise of interactive tokens/s. Release window noted as April 2025; always re-check the official source before buying hardware for a specific checkpoint.