robotnurse./gear/how-much-ram-for-local-llms — buyers' notes
How much RAM do you need for 7B, 13B, and 30B models?
Roughly: 7B models want about 8GB, 13B about 16GB, and 30B about 32GB of unified memory at q4, each with room for context — q8 roughly doubles each figure. 96GB+ buys comfort and headroom rather than capability you'll use every day.
Unified-memory budgets for 7B/13B/30B open-weight models at q4–q8: what fits in 32GB, 64GB, and why 96GB+ buys comfort, not just headroom.
At q4, a 7B model wants ~4–5GB of weights plus context; 13B ~8–10GB; 30B ~18–20GB. Leave headroom.
Unified memory is shared between the model and your workflow. Buy the step above what your model needs; the comfort costs far less than a second machine.
RELATED NOTES the full memory guide: RAM vs VRAM, context costs, offload math → · Mac Studio vs GPU PC — which box earns the desk → · three working builds under $3,000 →
CHECK PRICES see options at the retailer →
AFFILIATE DISCLOSURE: links in this post are affiliate links — if you buy through them I earn a small commission at no extra cost to you. RobotNurse is independent; I only recommend hardware I would actually deploy. No vendor paid for placement.
Picking the hardware is the easy half. I'll spec, install, and tune it end to end on your hardware.
start with the readiness audit → or book the call