Post

One formula that tells you exactly what fits on your GPU

If you’re running models locally, thinking “model → VRAM” falls apart once you account for how the weights were trained and quantized in the first place.

VRAM (in GB) ≈ Parameters (in billions) x (effective bits per weight ÷ 8)

This one formula explains everything across:

FP16 / BF16 → 16 bits → ~2 GB per 1B params

FP8 / INT8 → 8 bits → ~1 GB per 1B params