Skip to content
important Tools and apps price change

The cost of reading from the prompt cache for Claude Sonnet 5.5 has halved

clearly official source

According to the quoted announcement, the price of reading from the prompt cache for Claude Sonnet 5.5 has fallen from 0.20 to 0.10 USD per million tokens. The price of cache writes and all other prices remain unchanged.

The quoted announcement states that the price of reading from the prompt cache for Claude Sonnet 5.5 has fallen from 0.20 to 0.10 USD per million tokens. This is a reduction of 50 %, which applies to requests that read stored input from the cache.

The new rate is 0.05 times the base input price, down from 0.1 times previously. According to the announcement, the price of cache writes and all other prices remain unchanged. The quoted statement describes the change as already implemented but does not give a specific effective date.

What changed

Why it matters

With occasional use and a low volume of cache reads, savings will be small; in large-scale operations with frequent repeated reads, they will grow with the number of tokens billed at this rate. The halved rate applies only to cache reads, so overall savings depend on the share of costs accounted for by this item.

Relevant practical impact

What this means

01

For a business

Businesses running applications with Claude Sonnet 5.5 and a prompt cache now pay less for repeatedly reading stored input. Budgets must still account separately for cache writes and the other unchanged cost items.

Processes
What to decide Recalculate the cost of reading from the prompt cache based on the actual token volume at a rate of 0.10 USD per million tokens.
More business impacts →
Claude Sonnet 5.5 Prompt caching

Check the original

Event sources

clearly official source · 1 publisher, 0 independent. We count feeds from the same owner only once.

1
Claude Platform Release Notes primary source · first detected We've lowered the price of prompt cache reads on Claude Sonnet 5.5 from $0.20 USD to $0.10 USD per million tokens: 0.05x the base input price instead of 0.1x. Cache writes and all…