According to Vision Beating monitoring, the Claude Developer Console has launched the Prompt Caching Dashboard, accessible at platform.claude.com/usage/cache.
The dashboard is filtered by workspace, model, and time range, and mainly displays three key metrics: Cache Read Ratio, representing the percentage of requests that hit the existing cache; Cache Usage Breakdown, categorizing input tokens into Uncached, 5-Min Cache Write, 1-Hour Cache Write, and Cache Read, presented in a stacked bar chart format; Write Amortization, measuring how many times a single cache write has been reused by subsequent reads. In a screenshot example, Claude Opus 4.6 processed 2.74 billion input tokens over 7 days, with a read rate of 85.4% and a write amortization of 8.65x. At the bottom, there is a time series chart showing the cache read rate from 1 hour to 24 hours.
Anthropic's prompt caching mechanism allows API users to mark system prompts, long contexts, and other static content as cacheable. An extra fee is charged for the initial write, and subsequent hits are billed at approximately 10% of the standard input price. The cache is by default retained for 5 minutes, extendable to 1 hour with payment. Previously, users could only indirectly assess cache effectiveness through the token count field returned by the API, lacking a visualization tool.
