Anthropic Launches Claude Sonnet 5.5, Faster and Cheaper Than Its Predecessor

Anthropic has introduced Claude Sonnet 5.5, the second model in its Claude 5.5 family, positioning it as a faster and cheaper complement to the more powerful Claude Opus 5.5. The company says the new model runs more than 30% faster than Sonnet 5 and costs up to 30% less per task for most work, even though per-token pricing remains unchanged.

Sonnet 5.5 is designed for well-scoped everyday tasks: fixing bugs, building polished documents, slides and spreadsheets, and general-purpose coding. Opus 5.5 remains Anthropic's model for complex work requiring sustained judgment. A third model, Claude Haiku 5.5, aimed at high-volume and cost-sensitive applications, is expected to join the lineup in the coming weeks.

On benchmark performance, Anthropic reports dramatic gains. Sonnet 5.5 scored 70.6% on Terminal-Bench 4.0, an agentic coding evaluation, compared with just 10.3% for Sonnet 5. On GDPval-AA, which measures real-world work across 44 occupations, Sonnet 5.5 came within two points of Opus 5.5 and roughly 400 points ahead of its predecessor. The model also became the first Sonnet to beat the video game Pokémon Red using only screenshots, a marker Anthropic uses to illustrate long-horizon reasoning and image understanding.

At lower effort settings, Anthropic says Sonnet 5.5 can beat Sonnet 5's best score on several benchmarks for roughly a tenth of the cost per task. On coding-specific tests such as FrontierCode, it scored 10 points higher than Sonnet 5 at the same effort level while costing about one-fifteenth as much. On CursorBench, which draws on real coding sessions, its top score came within about two points of Opus 5.5.

Early corporate testers described the model favorably. Epic Games said Sonnet 5.5 cleared the same quality bar as higher-tier models in a system design audit and data flow review, managing tens of thousands of lines of gameplay code while staying responsive over multi-hour tasks. Slack reported that, without changing any prompts, Sonnet 5.5 outperformed Sonnet 5 on nearly all of its offline evaluations, using about 14% fewer output tokens.

Pricing for Sonnet 5.5 is unchanged from Sonnet 5: $2 per million input tokens, $10 per million output tokens, and $0.20 per million tokens for cache reads. Because the model typically needs fewer tokens to complete the same work, Anthropic says real-world costs fall by up to 30% per task.

On safety, Anthropic says its automated behavioral audit, covering roughly 1,850 scenarios, shows Sonnet 5.5 matching or improving on Sonnet 5 across most alignment and honesty measures, though Opus 5.5 still performs somewhat better overall. Because its cybersecurity capabilities are comparable to Opus 5's, Sonnet 5.5 is the first Sonnet model to launch with the same cyber safeguards and fallbacks used on Anthropic's most capable models; higher-risk cybersecurity tasks will be redirected to Sonnet 5. Biology safeguards remain the same as those on Sonnet 5, targeting a narrow set of high-risk requests while leaving routine research and clinical work largely unaffected.

The model is also the first Sonnet to launch with classifiers designed to prevent so-called distillation attacks, in which bad actors use large numbers of fake accounts to extract a model's capabilities without its built-in safeguards.

Claude Sonnet 5.5 is now available across major platforms, including Amazon Web Services, Google Cloud and Microsoft Azure, and through Anthropic's own Claude Platform under the identifier claude-sonnet-5-5. It is offered with zero data retention, matching the privacy terms of Opus 5.5 and Sonnet 5.