Pricing & Cost

Prompt Caching

Prompt caching stores encoded prompts for reuse. When the same context is used repeatedly with different queries, the context is reused from cache. Reused tokens cost 10-25% of standard price. Prompt caching requires minimum cache size (usually 1K tokens). Cache hits reduce cost and latency. Prompt caching is valuable for RAG, long documents, and multi-turn conversations. Implementation requires API support. Prompt caching is increasingly standard.