Sonnet 5.5(anthropic.com)
871 points by D2OQZG8l5BI1S06 15 days ago | 604 comments
tl;dr: Anthropic released Claude Sonnet 5.5, a faster (30%+) and cheaper (up to 30% less per task) upgrade to Sonnet 5, priced identically at $2/$10 per million input/output tokens. It shows major gains in agentic coding (70.6% vs 10.3% on Terminal-Bench 4.0) and approaches Opus 5.5 performance on several benchmarks while remaining a lower-cost complement for everyday tasks. It's also the first Sonnet model to ship with cybersecurity safeguards and anti-distillation classifiers, reflecting its significantly expanded capabilities.
HN Discussion:
  • ↓Opus 5.5 is sufficient and better, questioning Sonnet 5.5's value proposition
  • ↑Sonnet 5.5 excels at practical coding tasks like one-shotting a PacMan clone
  • ↑Sonnet 5.5's value is being the new free-tier default, bringing near-frontier to more users
  • ~Benchmark gains are misleading due to safeguard fallbacks skewing comparisons
  • ↓Chinese models offer far better price/performance, making Claude too expensive to justify