Model how much you save when a share of input tokens hits a published cached-input rate. Satellite of the main TokenCALC cost engine, not a second product.
Share of input tokens billed at the cached rate (~6,400 tokens).
No-cache request
$0.04768
With cache (80% hit)
$0.03168
Saved / request
$0.016 (33.6%)
Saved / month
$480.00
Steady-state math: cached input at the published cache rate; remaining input at standard input; output unchanged. Write cost (when published) is a one-time prefix charge. Repay it with the per-request savings above. Rates verified 2026-08-02.
Next step
Paste a real prompt in the main calculator with the same model and cache slider to combine Exact/Approx counts with this savings math.