Technical Trigger
The cache_control field has been added to the top level of the request body, allowing for automatic caching of prompts. The cache_control field can also be placed directly on individual content blocks for fine-grained control over what gets cached.
Developer / Implementation Hook
Developers can start using prompt caching by adding the cache_control field to their requests. For example, using the anthropic.Anthropic() client, developers can enable automatic caching by setting cache_control={"type": "ephemeral"}. Explicit caching can be achieved by placing cache_control directly on individual content blocks.
Structural Shift
The introduction of prompt caching represents a shift from processing entire prompts to reusing cached content, reducing processing time and costs.
Early Warning — Act Before Mainstream
To take advantage of this change, developers can:
* Start using automatic caching by adding the cache_control field to their requests
* Implement explicit caching by placing cache_control on individual content blocks
* Review the pricing structure for prompt caching, which introduces new multipliers for cache writes and reads