Claude Sonnet 5.5 Just Launched — And It's Nearly as Good as Anthropic's Flagship for Half the Effort

Anthropic has launched Claude Sonnet 5.5, the second model in its new 5.5 family, arriving less than a week after flagship Opus 5.5. The headline numbers: 30%+ faster output than Sonnet 5, and up to 30% lower cost per task — at the same list price of $2 per million input tokens and $10 per million output tokens.

The performance jump is the real story. On Terminal-Bench 4.0, an agentic coding evaluation, Sonnet 5.5 scored 70.6% — up from Sonnet 5's 10.3%, and just two points behind Opus 5.5 itself. On GDPval-AA, a benchmark spanning 44 occupations, Sonnet 5.5 scored 1844, nearly matching Opus 5.5's 1846. Anthropic says the cost reduction comes from the model needing fewer tokens to complete the same work, not from a price cut — a "smarter, not just cheaper" framing.

Sonnet 5.5 is also the first Sonnet-tier model to ship with the same cyber safeguards and reasoning-extraction protections previously reserved for Anthropic's most capable models, reflecting the company's assessment that its cybersecurity capabilities now rival Opus 5's.

It's live now on the Claude Platform, AWS, Google Cloud, and Microsoft Azure. A new Claude Haiku 5.5 is expected in the coming weeks to round out the family.

Bottom line: This lands the day before OpenAI's DevDay, and the timing isn't subtle — Anthropic is putting a strong, cheap, fast mid-tier model in front of developers right as OpenAI takes the stage. The benchmark gains are real and independently reported across multiple outlets. What's not yet independently tested is how Sonnet 5.5 performs against OpenAI's newest models head-to-head in production workloads — that comparison is still being made in real time by the developer community, not settled by either company's own numbers.