Why North Mini Code (free) pressures system RAM
North Mini Code (free) is Mixture-of-Experts: inference activates 3.8B active/token, but VRAM/RAM must usually hold the full ~30B expert set for fast routing. At Q4 the weight slab is ~16.9GB before KV (~0.01GB at 8K) and ~6GB OS/runtime overhead — totaling ~22.9GB raw, rounded to a 32GB kit. Stretching toward the full 256K-token window multiplies KV far faster than weights; that is the usual “I bought enough RAM for the model but still OOM” failure on Cohere MoE pages.
What RAM kit to buy
A 32GB dual-channel kit is enough for quantized North Mini Code (free) at modest context. Still prefer 2× matched SO-DIMM/UDIMM sticks; 1x RTX 3090 or RTX 4090 (24GB VRAM) covers the Flagship Consumer GPU GPU profile. If you chat with long pastes, jump a tier before the KV cache forces paging.
Workload notes
Cohere models like North Mini Code (free) are often enterprise-RAG oriented — size RAM for embedding caches and concurrent retrieval workers, not only the LLM weights. At 30B, North Mini Code (free) is a practical mid-size local model — sweet spot for single-GPU Q4/Q8 experimenters who still want headroom for IDE + Docker. Release window noted as 2025/2026; always re-check the model card before buying hardware for a specific checkpoint.