Home / Rank #2

Prompt caching

Typical savingsUp to 90% off cached input tokens
EffortLow — order your prompts, flip a flag

Prompt caching reuses the computed state behind a repeated prompt prefix (system prompt, tool definitions, long documents), so the static part of every request bills at a steep discount — up to 90% off on Anthropic with explicit cache breakpoints, and ~50% automatically on OpenAI.

Production reports commonly land in the 50–80% cache-hit range once prompts are structured with the stable content first and the variable content last. For chat apps, agents, and anything with a large system prompt, this is the highest-leverage one-day fix available.

How to do it

  1. Restructure prompts: stable content (system prompt, tools, reference docs) first, variable content (user message) last.
  2. On Anthropic, add cache_control breakpoints to the stable blocks; on OpenAI, caching applies automatically to repeated prefixes over 1024 tokens.
  3. Keep the prefix byte-identical across requests — any change above the breakpoint invalidates the cache.
  4. Track your cache-hit rate; below ~50% usually means something volatile (timestamps, request IDs) is leaking into the prefix.

Frequently asked questions

Does caching change the model output?

No. Prompt caching reuses computation for identical input prefixes; outputs are unaffected. It is purely a billing and latency optimization.

How long do caches live?

Provider-dependent: Anthropic offers 5-minute and 1-hour TTLs, OpenAI caches typically persist minutes and extend while in use. High-traffic prompts stay hot on their own.

Tools for this method

Open-source LLM gateway

LiteLLM

The default self-hosted gateway: one OpenAI-compatible proxy across 100+ providers with budgets, caching, routing, fallbacks, and …

Managed AI gateway

Portkey

Managed gateway with the strongest built-in semantic caching, plus guardrails, routing, and cost analytics. The low-ops route to t…

Next method: #3 Model routing & cascades