OpenAI Releases a Prompt Caching Dashboard Showing Hit Rate and Token Breakdown

On August 20, 2026, OpenAI released a Prompt Caching dashboard on its API platform. It tracks cache hit rate over time, cache reads per write, and the breakdown of cache-read, cache-write and uncached tokens, with filtering by model and service tier.

OpenAI API cost prompt-caching observability

On August 20, 2026, OpenAI released a Prompt Caching dashboard on its API platform¹. It surfaces cache hit rate over time, cache reads per write, and the breakdown of cache-read, cache-write and uncached tokens¹.

Metrics can be filtered by model and by service tier¹. OpenAI describes the purpose as understanding caching efficiency and identifying opportunities to improve it¹. Prompt caching bears directly on an API bill, yet how often it was actually hitting has been hard to track. The company also cut the price of its flagship GPT-5.6 Sol on August 21, so costs are moving on both the unit-price and the efficiency side.