Anthropic Releases Claude Sonnet 5.5: Frontier Coding at Mid-Tier Prices
Sonnet 5.5 scores 70.6% on the Terminal-Bench 4.0 agentic coding test — beating the more expensive Opus 5.5 — and ships with frontier-style cyber safeguards.
Anthropic released Claude Sonnet 5.5, the newest version of its mid-tier model, promising significantly faster performance at a lower cost than its predecessor.
The numbers are striking: 70.6% on the Terminal-Bench 4.0 agentic coding benchmark, up from 10.3% for Sonnet 5 and ahead of the 66.4% posted by the pricier Opus 5.5. It''s also the first Sonnet to ship with frontier-style cyber safeguards, including classifiers that block reasoning extraction.
The release lands in the middle of a frenzied model season — OpenAI recently shipped enhanced versions of its Sol and Luna models, and Meta announced a new model of its own.
For businesses, the practical impact is simple: the cost of capable AI keeps falling. Workloads that required a flagship model six months ago now run on the mid-tier — which means automation business cases that didn''t pencil out before may pencil out now.