Technology and Science22 September, 20262 min read
Anthropic Releases Claude Sonnet 5.5 With Faster, Lower-Cost Routine Task Performance
Anthropic released Claude Opus 5.5 and Claude Sonnet 5.5, expanding the 5.5 family. Opus 5.5 delivers Fable 5.1-level performance at about 40% lower running cost.
Claude Sonnet 5.5 Launch
Anthropic released Claude Sonnet 5.5 on Monday as an optimized alternative to Opus 5.5, positioning Sonnet 5.5 as faster and lower cost for routine tasks that still require substantial reasoning.
Anthropic said Sonnet 5.5 generates output more than 30% faster than Sonnet 5 and can cost up to 30% less per task, while keeping API input at $2 per million tokens and output at $10 per million tokens.
Anthropic also said Sonnet 5.5 is now generally available across platforms providing access to Anthropic models, and it described the release as a speed and security upgrade over Sonnet 5.
Anthropic’s lineup expanded within a week, with Opus 5.5 unveiled Sept. 22 and with Claude Haiku 5.5 described as coming in the coming weeks.
Pricing, Benchmarks, and Platforms
Anthropic priced Sonnet 5.5 at $2 per million input tokens plus $10 per million output tokens, and it said this is half the price of Opus 5.5. Anthropic said Sonnet 5.5 is live on Amazon Web Services, Google Cloud, and Microsoft Azure, and it described the model as improving coding, handling scoped tasks more reliably, and producing higher-quality documents, slide decks, and spreadsheets.
Anthropic reported a score of 1,811 on AA-Briefcase v1.1 and 61.6% on Chartography without tools, and it said Sonnet 5.5 is the first Sonnet version to beat "Pokémon Red" using screenshots alone. Theo Chu, a research product manager with Anthropic, told CNBC that "Sonnet is really for the cost-conscious customer where they might not need as much intelligence," and he added that Sonnet can fit routine tasks that need execution without Opus judgment.
Safety, Cyber Safeguards, and Fallout
Anthropic said Sonnet 5.5 does not advance the frontier of its models' capabilities, so most alignment testing focused on a "targeted set of risks that apply to models of any capability level." Anthropic also said Sonnet 5.5 has cybersecurity capabilities that are a "large improvement" over those of its predecessor, making it the first Sonnet model to launch with similar fallbacks and cyber safeguards to its more capable systems.
Anthropic said Sonnet 5.5 is the first Sonnet model to launch with classifiers designed to block distillation attacks, and it described the model as resisting efforts to extract it through distillation attacks. A federal appeals court upheld the government’s decision to exclude Anthropic from the Pentagon’s supply chain, and Anthropic said it was adding stronger cybersecurity safeguards alongside the performance gains.