How to cache conversation like OpenAI/Claude/DeepSeek?

The manual context caching’s content cannot be modified, If I want to update it, I need to create a new one, and the new one seems will be charged as no-cache-hit.

The default context caching seems cannot reduce cost

1 Like

Yes, having a way to do this becomes very important especially for building agents. We are currently not using Gemini 2.5 pro due to not having the option to cache increasing turns. Therefore in real use Sonnet 3.7 (90% caching price reduction) and o4-mini (75% caching price reduction) become much cheaper for agentic multi turn applications. I’d love to have a proper way to cache incremental conversation turns, which would solve this.

1 Like