Anthropic ships Claude Sonnet 5.5, 30% faster
The cheap tier just moved, but every number so far is Anthropic's own.
Anthropic has launched Claude Sonnet 5.5, the second model in the Claude 5.5 family. The company calls it a clear upgrade over Sonnet 5, says it runs more than 30% faster, and says it costs up to 30% less for most work.
It’s a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work.
That is the whole announcement. No independent founder benchmarks have been posted yet, no bills, no side by side runs against the model people are already paying for. Every figure in the launch is Anthropic's own.
Why the cheap tier matters more than the flagship
The frontier model gets the headlines. The cheaper, faster tier is what founders wire into a loop and run thousands of times, so a 30% price move there changes unit economics for anything run at volume, classifying, scoring, tagging, grading. Nobody has posted their own numbers on Sonnet 5.5 yet, so it is too early to say whether that plays out.
What to do about it
Nothing, for a week. "Up to 30% less" is a ceiling, not a rate, and "for most work" is doing quiet work in that sentence. Speed claims are similarly slippery once your prompts are long and your outputs are structured.
The useful move is to keep whatever eval you already trust, run it on Sonnet 5.5 against your current model with your real prompts, and look at your actual invoice at the end of the month. If the cost per task drops by a third on your workload, that is a real number and it is yours. Until then it is a marketing line from the vendor, and the vendor has every reason to round in its own favour.