Anthropic released Claude Sonnet 5.5 on Monday, positioning the new artificial-intelligence model as a faster and less costly option for everyday coding, document creation and other well-defined work. The company says the system generates output more than 30% faster than Claude Sonnet 5 and can complete many tasks at up to 30% lower cost, even though its listed token prices are unchanged. The launch fills out the middle of Anthropic’s new model family six days after the company introduced the more powerful Opus 5.5.
The company’s own results describe a sharp improvement in agentic coding and knowledge work. Anthropic reported a 70.6% score on Terminal-Bench 4.0 for Sonnet 5.5, compared with 10.3% for its predecessor, while its scores on workplace evaluations landed close to the larger Opus model. Anthropic also said the new system needed fewer tool calls and output tokens in tests, a combination that could matter to businesses running large numbers of automated tasks. Those figures are company-reported and will need independent replication across production workloads.
The release also extends safeguards that Anthropic previously reserved for its most capable systems. The company said Sonnet 5.5 is the first Sonnet model to ship with protections intended to block extraction of its reasoning through large-scale fake-account attacks. It also uses cyber controls similar to those on Opus 5.5, including fallbacks for a narrow set of higher-risk requests. Anthropic says ordinary software development should remain unaffected, but the change shows how stronger midtier systems are inheriting controls once associated mainly with flagship models.
Polymarket traders rapidly treated the release date as settled. The September 28 outcome in a contract on the next Claude Sonnet launch traded near a 99.9% implied probability after gaining about 54 percentage points over the latest day. The broader event recorded roughly $61,100 in activity over 24 hours. That price reflects the public launch rather than a forecast of how well the model will perform, and it should not be read as independent verification of Anthropic’s benchmark claims.
The next test will be adoption. Anthropic says Sonnet 5.5 is available across its own platform as well as Amazon Web Services, Google Cloud and Microsoft Azure, giving developers several routes to compare it with the earlier Sonnet release and with Opus 5.5. Buyers will be watching whether the promised speed and efficiency survive long coding sessions, repeated tool use and document-heavy workflows. External evaluations will also determine whether the striking jump on terminal tasks reflects a durable capability gain or sensitivity to a particular test setup. For Anthropic, the commercial wager is that a quicker, cheaper middle tier can carry more routine production work without requiring customers to pay for the flagship model.
