2026/03/30
What Is Prompt Caching TTL?
TTL is the lifetime of a prompt cache entry. Each hit refreshes it. Leave it unused for long enough, and it expires.
TTL stands for Time To Live: the amount of time a cache entry can survive without being used.
For prompt caching, TTL tells you how long a cached prefix can stay alive before the system discards it.
Start With a Concrete Example
A carton of milk has the same timing problem:
You buy milk with a shelf life of 5 days
Day 1 still fresh
Day 2 still usable
Day 5 last valid day
Day 6 expired and thrown awayPrompt cache TTL works the same way:
00:00 First request writes a prefix into cache
02:30 Another request hits the cache, TTL resets to 5 minutes
07:29 Another hit refreshes the timer again
12:30 No one uses it for 5 minutes, cache expires
12:31 A new request arrives, misses cache, and must rewrite it at full costTypical TTL Differences Across Providers
| Provider | TTL | Configurable | Refreshed on hit |
|---|---|---|---|
| Anthropic Claude | 5 minutes | No | Yes |
| OpenAI | Minutes to hours | No, opaque | Opaque |
| Google Gemini | Developer-defined, 1 hour by default | Yes | Yes |
If requests keep arriving, the cache stays alive. If usage pauses for too long, the entry expires and the next request has to write it again at full cost.
