OpenAI has introduced a new feature called prompt caching, which aims to reduce the costs and latency associated with input tokens. This feature is now available on TypingMind and applies automatically to the latest versions of GPT-4o, GPT-4o mini, o1-preview, and o1-mini models. Prompt caching provides a 50% discount on cached inputs and significantly reduces latency by 80% by reusing recently seen input tokens. The feature does not require any API changes and has a cache retention time of 5-10 minutes with a hard expiration of one hour. This innovation follows a similar feature released by AnthropicAI, which can save up to 90% on input tokens and reduce latency by 85%. The announcement was made during the OpenAI dev day.