Claude Sonnet 5.5 Targets Faster, Lower-Cost Everyday AI Work

Anthropic introduced Claude Sonnet 5.5 on September 28, 2026, calling it the second model in the Claude 5.5 family. [3] The company describes it as a clear upgrade over Sonnet 5 that runs more than 30% faster and costs up to 30% less for most work. [3]

Sonnet 5.5 retains Sonnet 5’s prices of $2 per million input tokens, $10 per million output tokens and $0.20 per million cache-read tokens. [3] Anthropic says the model typically requires far fewer tokens to complete the same work, producing costs up to 30% lower per task in its testing. [3]

Anthropic reports that Sonnet 5.5 scored 70.6% on the agentic coding evaluation Terminal-Bench 4.0, compared with 10.3% for Sonnet 5. [3] On GDPval-AA, which tests real-world work across different occupations, Sonnet 5.5 scored two points below the higher-end Opus 5.5. [3]

Anthropic describes Sonnet 5.5 as strongest at well-scoped everyday tasks, bug fixes and the creation of polished documents, slides and spreadsheets. [3] Anthropic research product manager Theo Chu told CNBC that Sonnet targets cost-conscious customers whose routine tasks require execution but not the judgment Opus can provide. [5]

Simon Willison reports that Sonnet 5.5 now powers Claude.ai’s free tier, while ChatGPT’s free tier uses Luna 5.6; he judges Anthropic’s free offering more capable. [1] In his free-tier test, Sonnet 5.5 produced a working WebGL page from a request for a three-dimensional pelican riding a bicycle. [1]

Anthropic says Sonnet 5.5 improves on or matches Sonnet 5 on most measures in its automated behavioral audit and launches with cyber safeguards comparable to those used for its most capable models. [3] Anthropic also says Haiku 5.5, intended for high-volume and cost-sensitive applications, will join the Claude 5.5 family in the coming weeks. [3]

Sonnet 5.5 creates a clearer price-performance ladder: Sonnet handles scoped execution at unchanged token rates, while Opus 5.5 remains positioned for work requiring more careful judgment. Making Sonnet 5.5 the free Claude.ai model also strengthens Anthropic’s entry-level offering relative to ChatGPT’s Luna 5.6, according to Willison’s assessment. The Terminal-Bench improvement supports Anthropic’s coding claims, although Willison’s separate maximum-effort failure shows why benchmark and efficiency gains still require workload-specific testing. What the sources don't address: how consistently the claimed per-task savings hold across production workloads, long contexts and different thinking-effort settings.