Opus 5.5 cuts every API price line: cache reads fall 60%, and an identical session bills about 31% less
Synopsis
Anthropic lowered the API list prices for Opus 5.5, the model used in Claude Code: input and output tokens cost 20% less than on Opus 5, and cache reads cost 60% less, with the read rate dropping from one tenth to one twentieth of the input price. What a task costs depends on turns, cache hits, output tokens and model price, because every turn resends the conversation so far. On an illustrative session with identical token counts, the price change alone takes the bill from $3.50 to $2.40, about 31% less; the stated 40% saving for typical workloads also assumes fewer tokens per task at the medium default.
Interpretation
Every API list price line for Opus 5.5 is lower than for Opus 5, and cache reads fall the most, from one tenth to one twentieth of the input price. Opus 5 billed cache reads at one tenth of the input price; Opus 5.5 bills them at one twentieth, while input and output tokens each drop 20%. An illustrative session with identical token counts prices at $3.50 on Opus 5 and $2.40 on Opus 5.5 at list prices, about 31% lower, a gap that comes from the price change alone.
The cost of a Claude Code task is set by four things, turns, cache reads, output token type and model, because each turn resends the conversation so far. The same conversation, which never grows past 120K tokens, accumulates about 2.8M input tokens over 40 turns and about 1.75M over 25 turns. For a context starting at 20K and growing to 120K, 40 turns with 90% cache hits cost about $1.62 in input and 25 turns about $1.02, while the same 2.8M input tokens with no cache cost $11.20.
On Opus 5.5 an output token costs 100 times a cache read, and thinking is billed as output, so changing effort moves the bill noticeably. The 60K output tokens of a typical task cost $1.20, the same as reading 6M tokens from cache. In the illustrative session output is only 60K tokens yet is the most expensive of the three lines at $1.50, against $1.00 for 2.0M cache reads.
Perspective
The figures are API list prices with illustrative token counts, excluding batch or volume discounts and cache writes; on Pro, Max and Team plans the same price cut shows up as limits going about 25% further rather than as a bill.
The stated 40% saving for typical workloads is an estimate that assumes Opus 5.5 uses fewer tokens per task at its medium default, so it is not a 40% cut in the price of a token; how much is saved varies with task shape, and the guidance is to measure it on your own work.