How much RAM does local AI need?
Estimate weight storage, context overhead, runtime requirements and safe system headroom.
Read the guide →Hardware
Match local models to real hardware.
Practical guidance for understanding memory, compute, bandwidth, power, thermals and the compromises behind local inference.
CoverageRAM, VRAM, CPUs, GPUs, laptops, desktops, workstations, servers, power consumption, thermals and hardware/model fit.
Published guide
Start with memory.
Key questions
Evaluate the complete system.
Can my computer run it?
Compare model size and context requirements with available RAM, VRAM and practical system headroom.
Where will inference run?
CPU-only inference, full GPU loading and partial GPU offload have different speed, memory and compatibility trade-offs.
Can the system sustain it?
Power limits, thermals, memory bandwidth and cooling affect sustained performance as well as operating cost.