Anthropic has launched Claude Sonnet 5.5, a new addition to its Claude model family positioned as a faster and lower-cost alternative to Claude Opus 5.5. The firm says Claude Sonnet 5.5 is created for well-scoped everyday tasks, software development, bug fixing and creating polished presentations and documents. The launch comes less than a week after Opus 5.5 and ahead of Haiku 5.5, which Anthropic says will arrive in the coming weeks. Sonnet 5.5 also brings significant gains over Sonnet 5 in coding, context, speed and efficiency.
How Claude Sonnet 5.5 Scores Against Sonnet 5 And Opus 5.5
Anthropic says Claude Sonnet 5.5 improves substantially over Sonnet 5 across multiple domains, especially agentic coding. On Terminal-Bench 4.0, an assessment of how well models complete complex, multi-step professional tasks through a command-line interface. Sonnet 5.5 scored 70.6%, compared with 10.3% for Sonnet 5.
The model also scored 52.1% on FrontierCode 1.1 Main at Xhigh effort, its primary reported figure, and lower at Max, compared with 42.4% for Sonnet 5 and 54.4% for Opus 5.5. On CursorBench 4.0, which assesses tasks drawn from real coding sessions. Sonnet 5.5 scored 55.5% compared with 34.1% for Sonnet 5 and 57.8% for Opus 5.5. Anthropic said Sonnet 5.5 also performs strongly on long-horizon tasks and image understanding. It is the first Sonnet model to beat Pokemon Red while working only from screenshots.
The model also improves in knowledge and context work. GDPval-AA v2.1, which tests real-world tasks across 44 occupations and nine sectors. Sonnet 5.5 scored 1,844 compared with 1,449 for Sonnet 5 and 1,846 for Opus 5.5. Anthropic said early testers found Sonnet 5.5 to be a more real collaboration partner than its predecessor. The company emphasized improvement in design, including the capability to polish user interfaces and follow presentation templates while creating slide decks. Speed is another major area. Claude Sonnet 5.5 generates outputs more than 30% faster than Sonnet 5, making it Anthropic’s fastest Sonnet model to date.
Why Sonnet 5.5 Costs Less Without A Price Cut
Sonnet 5.5 improves on Sonnet 5 across benchmarks, in some cases dramatically.
— Claude (@claudeai) September 28, 2026
It’s a faster, lower-cost complement to Claude Opus 5.5, strongest at well-scoped everyday tasks, fixing bugs, and creating polished documents, slides, and spreadsheets. pic.twitter.com/w8zB9WXxK7
Cost efficiency has also improved considerably with the new model. Claude Sonnet 5.5 has the same list pricing as Sonnet 5, which is $2 per million input tokens, $10 per million output tokens, and $0.20 per million tokens for cache reads. But Anthropic said the new model typically needs fewer tokens to complete the same work. Anthropic says Sonnet 5.5 can cost up to 30% less per task than Sonnet 5.
Why Claude Sonnet 5.5 Ships With Opus-Class Cyber Limits
Anthropic says Sonnet 5.5 does not advance the frontier of its model’s capabilities so its alignment assessment focused on targeted risks that apply across capability levels. On an automated behavioural audit covering 1,850 scenarios, the organization said Sonnet 5.5 improved on or matched Sonnet 5 on most measures involving alignment, resistance to misuse, and honesty. Anthropic also said the model came close to Opus 5.5 on nuance containment assessment, and was least likely of the internally tested models to probe the limits of its testing containment.
Anthropic acknowledged that assessments cannot reliably identify every possible failure mode. Cybersecurity is one area where the organization has extra safeguards. Anthropic said Anthropic says Sonnet 5.5’s cybersecurity capabilities are comparable to those of Opus 5, which is why it is the first Sonnet to ship with safeguards built for the company’s most capable models. Routine software development and bug fixing remain supported, while higher-risk cybersecurity tasks can fall back to Sonnet 5. The model also uses the same biology guardrails as Sonnet 5, with protections focused on harmful research. Anthropic said most research, education, and clinical work is unaffected, although some microbiology and virology requests may be flagged incorrectly.
Anthropic has not published false positive rates for the new cyber safeguards, nor an appeal route for work that is wrongly flagged. The AI company says routine software development bug fixing work remains supported, while higher-risk cybersecurity tasks can trigger a fallback to Sonnet 5. The organization’s material does not provide particular false positive rates or an appeal and tuning process for incorrectly flagged work.
Anthropic also said that Sonnet 5.5’s guardrails are narrow, and routine software development and most life sciences work are unaffected. The company’s approach combines behavioral testing with specific protections for cyber safety, biology, and classifiers that block attempts to extract the model’s reasoning. Claude’s Sonnet 5.5 therefore expands Anthropic’s mid-tier model lineup with a focus on coding, knowledge work, faster generation, and lower cost per task, while Opus 5.5 remains positioned for complex, open-ended work requiring sustained judgment. Anthropic is positioning Sonnet 5.5 for everyday professional workflows where speed is key.
Also Read: Claude AI Models Explained: Latest Claude 5, Claude 4 & Claude 3 Versions (2026)
