inference — Running a trained model to get answers — the phase where AI is actually used, as opposed to trained.
quantization — Shrinking a model by storing its numbers less precisely — like rounding — so it runs faster and fits on smaller hardware, at a small quality cost.
Why it matters
A 7GB FP4 build brings Kontext image editing within reach of consumer RTX cards that could not hold the 24GB model, at roughly twice the speed.
