Introducing Claude Sonnet 5.5
Anthropic says Sonnet 5.5 is over 30% faster than Sonnet 5 and can cost up to 30% less per task, while making a large jump on its Terminal-Bench 4.0 agentic-coding score. Its new system card also reports stronger prompt-injection resistance than prior Sonnet models, with cyber safeguards similar to Opus 5.5.
Treat the performance and cost figures as vendor-reported: the HN discussion questioned whether Sonnet is a better value than Opus at higher effort levels, and commenters noted that fallback behavior complicates some benchmark comparisons. Others saw a practical niche for faster, cheaper Sonnet agents handling well-scoped parallel tasks. Reactions were sharply split on the model’s cyber safeguards, with some reporting false positives on routine defensive work.