Hardware planner

Make room for your next model.

Calculate a baseline memory budget from model size and precision. No account required.

Your configuration

Use GPU VRAM, unified memory, or CPU system RAM. Do not add separate memory pools together.

Estimated baseline
4.73GiB

Model weights + runtime allowance

Model weights
3.73 GiB
Runtime allowance
1.0 GiB
Your budget after 15% reserve
20.4 GiB

Within your baseline memory budget

Context-window KV cache, image processing and runtime-specific allocations are additional. This is a lower-bound planning estimate, not a measured result or a guarantee that a model will run.

The calculation, in the open

Weights = parameters × one billion × bits ÷ 8 ÷ 1,073,741,824. Runtime allowance = the larger of 1 GiB or 20% of weight memory. We reserve 15% of your available memory for the system.