Which method is cheaper for specific doamin

which method is cheaper prompt,fine tunining or embedding

is it possible to achieve 100 percent accuarcy on fine tuned model for specific doamin