Claude Opus 5 Arrives as Anthropic's Summer Flagship

Anthropic has released Claude Opus 5, the flagship of its summer generation, at $5 per million input and $25 per million output tokens — setting the reference point that September's price cuts will be measured against.

Claude Opus 5 Arrives as Anthropic's Summer Flagship

Anthropic has released Claude Opus 5, the flagship tier of its 5-series generation, priced at $5 per million input tokens and $25 per million output tokens. The launch completes the summer lineup's top end, arriving a month after Sonnet 5 established the balanced tier at $2/$10.

Opus 5 inherits the mission its generation names imply: the hardest reasoning work in the catalog. Anthropic positions it for long-horizon coding, multi-document analysis, and agent orchestration — workloads where the premium over Sonnet 5 translates into fewer retries and less human cleanup, the arithmetic that justifies flagship pricing when it holds.

The pricing itself deserves attention as a reference point. At $5/$25, Opus 5 sits exactly where the previous flagship generation sat, continuing the industry pattern of holding flagship list prices steady while capability per dollar compounds. Watch this number through the rest of the year: with customer pressure on AI spending mounting and open-weight alternatives improving quarterly, the interesting question is not what Opus 5 costs today but what the next flagship will dare to charge.

For teams, the migration math is straightforward. Workloads already on Opus 4.x have a natural upgrade path with compatible APIs. Workloads on Sonnet 5 should stay there unless evaluation shows Opus 5 solving in one pass what Sonnet solves in three — the retry-rate arithmetic is where flagship models earn their price or fail to.

With the 5-series now complete from Haiku through Opus, attention turns to the fall — and to whether Anthropic's next moves extend the generation or replace it. On recent cadence, expect the answer before the leaves turn.