Claude Opus 5.5 Matches GPT-6 at 80% Less Cost

Anthropic put a number on its own progress this week: Claude Opus 5.5, the company says, matches the performance of its predecessor Claude Fable 5.1 on most tasks while costing 40 percent less to run. It's the first release in a new model family, with Sonnet 5.5 and Haiku 5.5 due in the coming weeks carrying similar claims about speed and cost.

The pitch is aimed squarely at the businesses paying Anthropic's bills. Input tokens drop to $4 per million from $5, output tokens fall to $20 from $25, a straight 20 percent price cut. Cache read costs drop 60 percent.

Anthropic says Opus 5.5 also needs fewer tokens to hit the same quality bar and produces output more than 30 percent faster than Opus 5, which pushes the real savings on total operating cost toward that 40 percent figure. Subscribers get a practical bonus too: five-hour usage windows stretch 20 percent, and with the lower per-token cost, Anthropic claims that budget now goes 25 percent further in practice.

Anthropic frames the release as a direct response to customer complaints, particularly from finance, law, and software teams, about cost, speed, and something else entirely: how the model actually writes.