How much does OpenAI prompt caching actually save?
Prompt caching reuses the computed prefix of a repeated prompt and bills it at a steep discount. Whether it's worth it depends entirely on how repetitive your prompts are — here's how to reason about it, and how to measure your own number.
The mechanic
Cached input tokens bill at a fraction of the standard input rate. The saving applies only to the portion of your prompt that is stable and repeated — typically a system prompt, instructions, or a document held constant across calls.
When it pays off
The bigger and more repeated your prefix, and the more calls share it, the larger the win. Short, highly-variable prompts benefit little; long shared context benefits a lot.
Turn this into your number
Drop your usage export into the free analyzer and see how much of this applies to your account — provable savings separated from estimates. Nothing is uploaded.
Analyze my usage — free