Google Introduces Implicit Caching for Gemini AI Models

Google has introduced a new feature called 'implicit caching' in its Gemini API, which aims to reduce costs for developers using its latest AI models. This feature is designed to automatically enable up to 75% savings on repetitive context passed to models, specifically benefiting the Gemini 2.5 Pro and 2.5 Flash models.

Implicit caching differs from Google's previous explicit caching method by automating the process, thus eliminating the need for developers to manually define high-frequency prompts. This change comes after feedback from developers who found the explicit caching method cumbersome and costly.

The new caching system is activated by default for Gemini 2.5 models. It works by identifying requests that share a common prefix with previous ones, allowing them to hit a cache and pass on cost savings. The minimum prompt token count required for a cache hit is 1,024 for 2.5 Flash and 2,048 for 2.5 Pro models.