How this calculator works
Enter model parameters (billions), bits per parameter, runtime overhead (%). Select Calculate to apply the displayed formula and review the labeled results.
Formula / method
Memory bytes = parameters × 1,000,000,000 × bits ÷ 8 × (1 + overhead ÷ 100)
Worked example
A 7-billion-parameter model at 16-bit precision with 20% overhead needs about 16.8 GB.
Assumptions and limitations
Every parameter uses the selected bit width and overhead is applied uniformly.
Actual inference memory also depends on KV cache, activations, framework buffers, quantization metadata, batch size and device allocation.