Why Codestral 22B pressures system RAM
Codestral 22B is a dense 22B network — every weight participates each token, so quantization choice dominates. Q4_K_M lands near ~12.4GB weights, plus ~0.05GB KV at 8K and ~6GB overhead (~18.5GB → 32GB kit). The 33K-token context ceiling is the sleeper cost: long-doc or agent traces inflate KV while the 22B slab stays fixed. Prefer dual-channel DDR5 bandwidth when CPU offload or mmap is involved.
What RAM kit to buy
A 32GB dual-channel kit is enough for quantized Codestral 22B at modest context. Still prefer 2× matched SO-DIMM/UDIMM sticks; 1x RTX 4060 Ti (16GB) or RTX 4070 Ti Super (16GB) covers the 16GB VRAM Single GPU GPU profile. If you chat with long pastes, jump a tier before the KV cache forces paging.
Workload notes
Mistral releases like Codestral 22B are common in production vLLM; dual-channel bandwidth helps prompt throughput when CPU offload is in play. At 22B, Codestral 22B 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 May 2024; always re-check the official source before buying hardware for a specific checkpoint.