Accurate GGUF Memory Calculator

Estimate memory usage for GGUF models based on GPU layers, context length, and cache type.

The formula was discovered through symbolic regression using TuringBot, evaluating over a billion candidate formulas against 19,517 real measurements. For details, see this blog post.

GGUF Model URL

GPU Layers

--gpu-layers in llama.cpp.

0 256

Context Length

--ctx-size in llama.cpp.

512 262144

Cache Type

Cache quantization.

fp16 q8_0 q4_0

Estimated memory usage:

Status