Anthropic launches Claude Sonnet 5.5

Anthropic on Monday released Claude Sonnet 5.5, the second model in its Claude 5.5 family, saying the mid-tier model runs more than 30% faster than Sonnet 5, costs up to 30% less per task at the same $2/$10 per-million-token list price, and ships with cyber safeguards previously reserved for its most capable models.

Anthropic just dropped a faster, cheaper mid-tier Claude for daily coding and office work, and paired the capability jump with cyber safeguards it used to reserve for its top models.

On Monday, 28 September 2026, Anthropic introduced Claude Sonnet 5.5 as the second model in the Claude 5.5 family. The page is anthropic.com/claude-sonnet-5-5, dated September 28, 2026. Anthropic calls it a clear upgrade over Claude Sonnet 5. In this lineup, Opus is the most capable model, Sonnet is the mid-tier model, and Haiku is the smaller one built for high volume. These lines are Anthropic’s.

Speed and the bill, still the company’s. Anthropic says Sonnet 5.5 generates output more than 30% faster than Sonnet 5, and that it is the fastest Sonnet the company has shipped. It also says the model costs up to 30% less per task for most work. A token is a small chunk of text the model reads or writes. The list price is unchanged from Sonnet 5: $2 per million input tokens, the text you send, and $10 per million output tokens, the text the model writes. A cache read, a saved piece of a long prompt you do not pay full price for again, is $0.20 per million tokens. Anthropic says the lower bill comes from using fewer tokens to finish the same task, not from a cheaper sticker price.

What the company says it is for. Anthropic calls Sonnet 5.5 a faster, lower-cost complement to Claude Opus 5.5. Opus 5.5, on the same page, is for complex work that needs careful judgment. Sonnet 5.5 is strongest, Anthropic says, at well-scoped everyday tasks, fixing bugs, and polished documents, slides, and spreadsheets. Well-scoped means the task already has a clear boundary.

The coding scores are Anthropic’s evaluations. They are not a rerun by an outside lab. Terminal-Bench 4.0 is a test of multi-step coding in a command-line terminal, the text window where a program takes a series of steps instead of one reply. Anthropic reports 70.6% for Sonnet 5.5 and 10.3% for Sonnet 5. CursorBench 4.0 uses ambiguous, multi-file tasks taken from real sessions in the Cursor code editor. Anthropic reports 55.5% for Sonnet 5.5 and 34.1% for Sonnet 5. A higher percent means more of those test tasks were completed. On the same tables, Anthropic prints Opus 5.5 at 66.4% on Terminal-Bench 4.0 and 57.8% on CursorBench 4.0. The company also says that in its own testing, and in testing by outside groups, Opus 5.5 remains clearly stronger at complex, open-ended work that needs sustained judgment. A benchmark score is one slice of what a model can do.

Where it is on. Anthropic says Sonnet 5.5 is available now on the Claude Platform, Amazon Web Services, Google Cloud, and Microsoft Azure. On the Claude Platform the model’s name is claude-sonnet-5-5. The Claude Platform is the service a developer calls. Anthropic says Sonnet 5.5 is available with zero data retention, the same as Opus 5.5 and Sonnet 5.

Cyber and biology, as Anthropic describes the limits. The company says Sonnet 5.5’s cybersecurity skills are comparable to Opus 5’s, an earlier top model. Because of that jump, Anthropic says this is the first Sonnet to launch with cyber safeguards and fallbacks like the ones on its most capable models. A safeguard is a limit meant to stop misuse. Anthropic says higher-risk cybersecurity tasks will visibly fall back to Sonnet 5, while people can still find and fix bugs in their own code as ordinary software work. Biology safeguards, the company says, are the same set used on Sonnet 5. Both sets of limits target a narrow group of high-risk requests. Anthropic says routine software development and most life-sciences work are unaffected, though some microbiology and virology requests may be flagged by mistake. The company also says people who defend networks will soon be able to apply to an expanded Cyber Verification Program for tiered access on Sonnet 5.5, Opus 5.5, and Claude Mythos models. That wider access is a plan on the page.

What is still ahead. Anthropic says Claude Haiku 5.5 will join the Claude 5.5 family in the coming weeks. Haiku 5.5 is the model aimed at high-volume, cost-sensitive work. “Coming weeks” is a schedule, not a launch that has happened.

Reuters reported the same Monday that Anthropic released Sonnet 5.5, the second model in the Claude 5.5 family, as the lab expands its lineup ahead of a planned initial public offering. An initial public offering is a first sale of shares to the public. Reuters does not print a date for that sale, a price for the company, or a revenue figure. Reuters repeats the $2 and $10 prices, says Anthropic described the model as a faster, lower-cost complement to Opus 5.5, and says Opus 5.5 launched last week at $4 per million input tokens and $20 per million output tokens.

The picture is the cover of Anthropic’s System Card for Claude Sonnet 5.5. The page reads “System Card: Claude Sonnet 5.5” and “September 28, 2026.” It is the document, not a photograph of a product.