OpenAI enhances prompt caching for gpt-6 to boost efficiency

OpenAI has introduced an improved prompt caching system for GPT-6, designed to increase cache hit rates and reduce costs for developers. The update allows for cache discounts on shared prefixes reused within a 30-minute window, with potential savings of up to 90% on cached input tokens. Developers can now monitor cache performance through a new dashboard, which tracks hit rates and compares cached versus uncached tokens. The system also includes diagnostic tools to investigate unexpected cache misses, helping developers identify changes that may have disrupted reuse. OpenAI encourages developers to use explicit cache breakpoints to control which prompt prefixes are reused and to adjust reasoning effort without breaking the cache. Prewarming features are also available to reduce latency by preparing context ahead of time. These enhancements aim to optimize caching for various workloads, with additional guidance provided on managing tool changes and improving setup through the prompt caching guide.