Anthropic unveils Claude Sonnet 5.5
- Maxime Hiez
- Anthropic
- 30 Sep, 2026
Introduction
Anthropic announces Claude Sonnet 5.5, the second model in the Claude 5.5 family after Opus 5.5. The model generates responses more than 30% faster than Sonnet 5 and costs up to 30% less per task, at unchanged pricing.
Check the Claude Opus 5.5 article HERE.
Positioning within the 5.5 lineup
Where Opus 5.5 targets complex work requiring sustained judgment, Sonnet 5.5 excels at well-scoped everyday tasks : fixing bugs, producing polished documents, presentations, and spreadsheets. Anthropic also credits it with a sharp eye for visual detail. Claude Haiku 5.5, aimed at high-volume, cost-sensitive use cases, will join the family in the coming weeks.
Performance
On reference evaluations :
| Sonnet 5.5 | Sonnet 5 | Opus 5.5 | |
|---|---|---|---|
| Terminal-Bench 4.0 | 70.6% | 10.3% | 66.4% |
| CursorBench 4.0 | 55.5% | 34.1% | 57.8% |
| GDPval-AA v2.1 | 1844 | 1449 | 1846 |
| Humanity’s Last Exam | 64.5% | 54.9% | 67.7% |
| OSWorld 2.1 (partial) | 80.1% | 57.0% | 81.8% |
On several evaluations, Sonnet 5.5 at Low or Medium effort beats Sonnet 5’s best score for about a tenth of the cost per task. At higher effort, it reaches performance comparable to Opus 5.5’s. Anthropic tempers this, however : on complex, open-ended work requiring sustained judgment, Opus 5.5 remains clearly stronger, both in its own testing and in that of external testers.
Coding
Sonnet 5.5’s performance jump is particularly clear in coding. At High effort on FrontierCode, it scores 10 points higher than Sonnet 5 at the same effort level, for about a fifteenth of the cost per task. On CursorBench, its best score sits within about two points of Opus 5.5’s.
- Epic Games : Reports that Sonnet 5.5 “cleared the same quality bar you’d expect from a higher-tier model”, managing tens of thousands of lines of gameplay architecture code with less prescriptive prompting.
- Lovable : Reports a third fewer tool calls and roughly half the shell runs to finish a task, compared with Sonnet 5.
Knowledge work
On GDPval-AA, which tests real-world tasks across 44 occupations and nine industries, Sonnet 5.5 scores nearly level with Opus 5.5, about 400 points above Sonnet 5.
- Balyasny Asset Management : Reports a higher score than Sonnet 5 on a suite of 2,441 finance tasks, using about 121,000 tokens per answer compared with 497,000 for Sonnet 5.
- Box : States that Sonnet 5.5 rechecks data in source documents and catches errors Sonnet 5 missed, while being 2.4 times faster and using 12% fewer tokens.
Alignment and safeguards
Sonnet 5.5 doesn’t extend the Claude lineup’s capability frontier ; the alignment assessment therefore focused on a targeted set of risks applicable at any capability level. On the automated behavioral audit, which covers about 1,850 scenarios, Sonnet 5.5 matches or improves on Sonnet 5 on most measures of alignment, resistance to misuse, and honesty.
- Cybersecurity : Sonnet 5.5’s cyber capabilities improve significantly over Sonnet 5’s, earning it safeguards similar to Opus 5.5’s. High-risk cybersecurity tasks are redirected to Sonnet 5.
- Biology : Sonnet 5.5 keeps the same safeguards as Sonnet 5, targeting harmful requests without affecting most research, education, and clinical work.
- Anti-distillation : The first Sonnet model to ship with safety classifiers that block reasoning extraction, and to extend the preserved thinking mechanism already deployed on Fable 5.1 and Opus 5.5.
Cost and availability
Sonnet 5.5’s pricing stays identical to Sonnet 5’s :
- Input : 2$ / 1M tokens
- Output : 10$ / 1M tokens
- Cache (read) : 0.20$ / 1M tokens
- Cache (write) : 2.50$ / 1M tokens
info
Claude Sonnet 5.5 is available now on all platforms, including Amazon Web Services, Google Cloud, and Microsoft Azure. On the Claude Platform, developers can access it under the identifier claude-sonnet-5-5. The model remains available with zero data retention.
warning
Conclusion
Sonnet 5.5 isn’t trying to push the Claude lineup’s capability frontier, but to bring Sonnet’s price-to-performance ratio closer to the level reached by Opus 5.5, at unchanged pricing. For well-scoped agentic workflows, from code to documents to customer support, it’s now the lineup’s most rational entry point, with Opus 5.5 reserved for work that demands sustained judgment over time.
Sources
Did you enjoy this post ? If you have any questions, comments or suggestions, please feel free to send me a message from the contact form.
Don’t forget to follow us and share this post.
