Anthropic ships Sonnet 5.5 and holds the sharper cyber use back
The model is faster and cheaper on ordinary work. Higher-risk cyber tasks fall back to Sonnet 5 unless a defender is verified.
SAN FRANCISCO — Anthropic released Claude Sonnet 5.5 on 28 September. The company said it is more than 30 percent faster than Sonnet 5 and costs up to 30 percent less for most work, because it uses fewer tokens. The list price is unchanged from Sonnet 5, at 2 dollars per million input tokens and 10 dollars per million output tokens.
Sonnet 5.5 is positioned under Opus 5.5. Anthropic said it is strongest at scoped everyday tasks, bug fixes, and documents, slides and spreadsheets. It is on the Claude platform as claude-sonnet-5-5, and on Amazon Web Services, Google Cloud and Microsoft Azure. Haiku 5.5 is still to come.
The cyber line is the one a security team should read. Anthropic said Sonnet 5.5’s cyber capability is a large step up, so higher-risk cyber tasks fall back to Sonnet 5. Routine bug fixing in a team’s own code still runs. A wider Cyber Verification Program, for tiered access on Sonnet 5.5, Opus 5.5 and Claude Mythos, is described as coming soon, not as open today.
A buyer who turns the new model on for a general assistant has not turned on the cyber tier. Those are different permissions. 54east’s governance note for public institutions is the parallel question. https://54east.ai/insights/ai-governance-gap-uae-public-institutions.html
