Source: Gemini API Context Caching and Pricing Docs

Google Gemini API documentation on implicit and explicit context caching, plus the pricing page's cache and batch rates.

Contents
  • What: official API documentation (caching guide and pricing page)
  • Author/publisher: Google
  • Date: living documents; read 2026-11-14 via the AI Studio docs bundle
  • Original: Context caching

Summary

Gemini offers two caching modes. Implicit caching is automatic on Gemini 2.5 and newer and passes savings through when a request happens to hit. Explicit caching (generateContent API only, beta, not supported on Gemini 4) creates a named cachedContents resource with a TTL that you reference in later calls, which guarantees the discount but adds an hourly storage charge. The pricing page lists cached-token and storage rates per model, including Batch-tier rates.

Key claims

  • Implicit caching is enabled by default for Gemini 2.5+ with no cost-saving guarantee; explicit caching guarantees savings and is available on Gemini 3 and 2.5 models only.
  • Minimum tokens for a cache: 4,096 on Gemini 4 models and Gemini 3.x; 2,048 on Gemini 2.5 Flash/Pro.
  • Hit rates improve when large shared content sits at the start of the prompt and similar-prefix requests arrive close together in time.
  • On Gemini 4 multi-turn/agent loops, the whole previous request is cached only if the new request adds at least 2,050 tokens; otherwise the last 2,050 tokens are reprocessed at full price.
  • Explicit caches default to a 1-hour TTL; ttl or expire_time can be set and later updated, nothing else about a cache can change.
  • Pricing examples (paid tier, through end of 2026): Gemini 3.8 Flash input $0.75/MTok, cached $0.075/MTok, storage $0.50 per MTok per hour; Gemini 4 Carbon input $2.00, cached $0.10, implicit only. Gemini 4 Carbon prices double on 1 Dec 2026; Gemini 3.8 Flash prices (including storage) double on 1 Jan 2027.
  • Cache hits are reported in usage metadata (total_cached_tokens in the Interactions API).

Feeds

Prompt caching · Batch processing

Suggest changes →