Skip to content

Claude Sonnet 5.5's prompt-cache reads get 50% cheaper

Anthropic cut the cache-read price for Sonnet 5.5 from $0.20 to $0.10 per million tokens, a change it says lowers agentic workload costs by around 20%.

Source: Anthropic

Anthropic cut the price of prompt-cache reads on Claude Sonnet 5.5, effective 7 October. A cache hit now costs $0.10 per million tokens, down from $0.20 — moving from 0.1x the base input price to 0.05x, the same multiplier Opus 5.5 already uses. Cache writes and every other Sonnet 5.5 price are unchanged.

The cut landed the same day as the Claude Haiku 5.5 launch, but it’s a separate change aimed at a different cost line. Anthropic says it lowers the cost of running Sonnet 5.5 on most agentic tasks by roughly 20%, because cache reads make up a large share of the tokens an agent burns through when it keeps resending the same system prompt, tool definitions or conversation history.

Why it matters: cache reads, not writes, dominate token spend in anything that repeatedly replays context — coding agents, RAG pipelines, long multi-turn chats. Halving that rate moves the economics for Sonnet-5.5-based agents without requiring any code change beyond the cache_control block you’ve presumably already set.

"cache_control": { "type": "ephemeral" }
Share

Related reading

More from The Wire

All briefs

Anthropic cuts distillation attacks on Claude by up to 99%

New safeguards, including a reasoning classifier and stricter account limits, now block most attempts to copy Claude by training on its output. Blocked requests in high-confidence cases are billed, and some integrations may need updating.

via Anthropic