Anthropic introduces Claude Sonnet 5.5 — a faster, lower‑cost Sonnet for everyday developer tasks
Anthropic has announced Claude Sonnet 5.5, the second model in its Claude 5.5 family, which the company says is a faster, lower‑cost complement to Opus 5.5 and is tuned for well‑scoped everyday tasks such as bug fixes, polished documents, slides, spreadsheets, and design work. [1]
What Sonnet 5.5 is designed to do
Anthropic describes Sonnet 5.5 as intended for fast iteration on less complex tasks, while Opus 5.5 is positioned for complex work requiring careful judgment; Sonnet 5.5 is also claimed to excel at image understanding and long‑horizon work relative to prior Sonnet models. [1]
Performance and benchmarks
Anthropic reports Sonnet 5.5 achieves 70.6% on Terminal‑Bench 4.0 (an agentic terminal coding evaluation) versus Sonnet 5’s 10.3%, and provides a set of comparative scores across FrontierCode, CursorBench, GDPval‑AA, and Chartography where Sonnet 5.5 shows substantial gains over Sonnet 5 and in many cases approaches Opus 5.5’s results. [1]
Below is a compact summary of the prices, availability, and a few headline benchmark numbers Anthropic published:
| Item | Value |
|---|---:|
| Announcement date | 2026‑09‑28 |
| Input tokens | $2 / million |
| Output tokens | $10 / million |
| Cache reads | $0.20 / million |
| Terminal‑Bench 4.0 | 70.6% |
| CursorBench 4.0 | 55.5% |
| GDPval‑AA | 1844 |
[1]
Anthropic also provides cost‑vs‑accuracy plots and reports that at Low or Medium effort Sonnet 5.5 often exceeds Sonnet 5’s best scores for a fraction of the cost per task, and states that at higher effort it can perform comparably to Opus 5.5 at similar cost levels. [1]
Speed, cost, and safety tradeoffs
Anthropic claims Sonnet 5.5 generates outputs more than 30% faster than Sonnet 5 and typically requires far fewer tokens to complete the same work, leading to claims of up to 30% lower cost per task in the company’s testing. [1]
On safety, Anthropic reports Sonnet 5.5 improves on or matches Sonnet 5 on most automated alignment measures, and—because its cybersecurity capabilities are comparable to Opus 5—Sonnet 5.5 launches with cyber safeguards and fallbacks similar to those used for Anthropic’s more capable models; Anthropic says its biology safeguards remain the same as Sonnet 5’s and are targeted to a narrow set of high‑risk requests. [1]
Developer impact and early tester reports
Anthropic includes a set of early tester quotes and reported metrics: testers described Sonnet 5.5 as quicker to iterate with, more efficient in token use, and better at understanding codebases; one reported that Sonnet 5.5 batched tool calls more than Sonnet 5, reducing steps and cost. Companies including Epic Games, CodeRabbit, SpaceXAI, Unity, and others are quoted on real‑world coding and app‑build experiments presented by Anthropic. [1]
For teams considering integration, Anthropic frames Sonnet 5.5 as an option when the priority is fast, cost‑efficient handling of well‑scoped software and content workflows, with Opus 5.5 reserved for higher‑risk or more complex judgment tasks. [1]