Anthropic has launched Claude Sonnet 5.5, the latest model in its Claude 5.5 family, with major improvements in coding, speed and cost efficiency. Announced on September 28, 2026, Sonnet 5.5 is designed for well-scoped everyday work, software development, bug fixing and creating documents, presentations and spreadsheets.
Anthropic says Claude Sonnet 5.5 generates responses more than 30% faster than Sonnet 5 and can cost up to 30% less per task because it typically uses fewer tokens to complete the same work. Its listed token prices, however, remain unchanged from Sonnet 5.
What Is Claude Sonnet 5.5?
Claude Sonnet 5.5 is the second model in Anthropic’s Claude 5.5 generation, following Claude Opus 5.5.
Anthropic positions Sonnet 5.5 as a faster and more cost-efficient model for tasks where the overall approach is relatively clear and the model needs to execute the work efficiently. This includes coding, debugging, knowledge work and creating polished business content.
The model is intended to complement Opus 5.5 rather than replace it. Anthropic describes Opus 5.5 as being better suited to complex work requiring sustained judgment, while Sonnet 5.5 is aimed at faster execution of well-defined tasks.
Major Coding Improvements
Coding is one of the areas where Anthropic reports the largest improvement.
On Terminal-Bench 4.0, an evaluation of agentic coding and complex command-line tasks, Sonnet 5.5 scored 70.6%, compared with 10.3% for Sonnet 5 in Anthropic’s testing.
On CursorBench 4.0, which evaluates coding agents on ambiguous, multi-file tasks based on real Cursor sessions, Sonnet 5.5 scored 55.5%, compared with 34.1% for Sonnet 5. Anthropic says the model’s best result was within roughly two percentage points of Opus 5.5 on that benchmark.
Anthropic also reports that at High effort on FrontierCode 1.1, Sonnet 5.5 scored 10 percentage points higher than Sonnet 5 at the same setting while requiring substantially less cost per task.
The company says early testers noticed that Sonnet 5.5 can understand codebases quickly and batch tool calls more effectively, reducing the number of steps needed to complete coding tasks.
Faster Responses Without a Higher Token Price
One of Sonnet 5.5’s biggest changes is speed.
Anthropic says the model generates output 30% or more faster than Sonnet 5, making it the fastest Sonnet model the company has released.
The listed API pricing remains:
| Pricing | Claude Sonnet 5.5 |
|---|---|
| Input tokens | $2 per million |
| Output tokens | $10 per million |
| Cache reads | $0.20 per million |
| Cache writes | $2.50 per million |
The important distinction is between token pricing and cost per completed task.
Anthropic has not reduced the headline input and output token prices compared with Sonnet 5. Instead, it says Sonnet 5.5 generally requires fewer tokens to accomplish the same work. As a result, Anthropic reports that the model can cost up to 30% less per task than Sonnet 5.
How Sonnet 5.5 Compares With Sonnet 5
Anthropic reports substantial gains across several evaluations.
| Evaluation | Claude Sonnet 5.5 | Claude Sonnet 5 |
|---|---|---|
| Terminal-Bench 4.0 | 70.6% | 10.3% |
| FrontierCode 1.1 | 46.2% at Max effort | 42.4% |
| CursorBench 4.0 | 55.5% | 34.1% |
| GDPval-AA v2.1 | 1844 | 1449 |
| AA-Briefcase v1.1 | 1811 | 1359 |
| Humanity’s Last Exam | 64.5% with tools | 54.9% with tools |
| OSWorld 2.1 | 80.1% partial | 57.0% partial |
These figures come from Anthropic’s published evaluations, and the company notes that benchmark results represent only one aspect of model performance.
Anthropic also notes that some evaluations have different effort settings and methodologies, so scores should not be treated as direct measurements of every real-world workload.
Sonnet 5.5 Targets Everyday AI Work
Although coding receives significant attention in the launch, Sonnet 5.5 is not limited to software development.
Anthropic says the model has improved across knowledge-work tasks and can create more polished documents, spreadsheets and presentations.
On GDPval-AA, which evaluates real-world work across 44 occupations and nine industries, Sonnet 5.5 achieved a score of 1844, compared with 1449 for Sonnet 5 and 1846 for Opus 5.5 in Anthropic’s published results.
The model also achieved an 1811 score on AA-Briefcase, compared with 1359 for Sonnet 5. Anthropic describes AA-Briefcase as an evaluation of long-horizon knowledge work.
Effort Settings Let Users Trade Speed for Quality
Sonnet 5.5 includes configurable effort levels that allow users to balance speed, cost and the amount of reasoning used.
Anthropic says the default effort is Medium in Claude apps and Claude Code, while the Claude Platform defaults to High. Lower settings generally allow faster and less expensive responses, while higher settings give the model more time to reason and check its work.
This means the same model can be used differently depending on the task.
A developer working on a routine bug fix may prefer a lower effort setting, while a more complicated coding or research task can use higher effort when additional reasoning is useful.
New Safety Measures for Cybersecurity
The coding improvements also come with additional safeguards.
Anthropic says Sonnet 5.5’s cybersecurity capabilities have improved significantly compared with Sonnet 5. Because of that, the model launches with safeguards and fallback mechanisms similar to those used with Opus 5.5.
Routine software development and normal bug fixing remain supported, while higher-risk cybersecurity requests can fall back to Sonnet 5.
Anthropic also says Sonnet 5.5 uses new safety classifiers designed to prevent reasoning extraction through large-scale distillation attacks.
Claude Sonnet 5.5 Availability
Claude Sonnet 5.5 is available across Anthropic’s supported platforms, including Amazon Web Services, Google Cloud and Microsoft Azure.
Developers can access the model through the Claude Platform using the model identifier:
claude-sonnet-5-5
Anthropic also says the model is available with zero data retention, consistent with its stated availability for Opus 5.5 and Sonnet 5.
What the Launch Means for Developers
The main change with Sonnet 5.5 is not simply a higher benchmark score.
For developers, the combination of stronger coding performance, faster output and lower reported cost per task could make the model more practical for workloads that run repeatedly.
That matters particularly for coding agents, where every additional tool call, generated token and iteration can add both time and expense.
Anthropic says early testing showed Sonnet 5.5 completing coding tasks with fewer steps and tool calls than Sonnet 5. The company also cites feedback from organizations including Epic Games, Unity, CodeRabbit and Base44 describing improvements in coding workflows.
These are company-provided or early-tester results rather than independent evidence across every development environment, so actual performance can vary depending on the task, tools and configuration.
Sonnet 5.5 vs Opus 5.5
Anthropic is positioning the two models for different workloads.
Sonnet 5.5 is focused on faster, well-scoped work, while Opus 5.5 is intended for more complex tasks that require sustained judgment.
On some benchmarks, Sonnet 5.5 approaches Opus 5.5, but Anthropic explicitly says Opus remains stronger for complex, open-ended work requiring sustained judgment.
This gives developers a choice between a more cost-efficient model for routine and moderately complex work and a higher-tier model for demanding tasks.
What Happens Next?
Anthropic is continuing to expand the Claude 5.5 family. The company says Claude Haiku 5.5, designed for high-volume and cost-sensitive applications, will join the family in the coming weeks.
For Sonnet 5.5, the immediate focus will be adoption across coding tools, enterprise applications and other agentic workflows.
The key question will be how its reported gains translate into real-world workloads outside Anthropic’s published evaluations. The model’s lower token usage, faster output and improved coding performance could make it attractive for developers running AI agents at scale, but actual cost and performance will depend on task complexity, effort settings and usage patterns.
Frequently Asked Questions
What is Claude Sonnet 5.5?
Claude Sonnet 5.5 is Anthropic’s latest Sonnet model, designed for coding, everyday knowledge work, bug fixing and creating documents, presentations and spreadsheets.
How much does Claude Sonnet 5.5 cost?
The model is priced at $2 per million input tokens, $10 per million output tokens, $0.20 per million cache-read tokens and $2.50 per million cache-write tokens.
Is Claude Sonnet 5.5 cheaper than Sonnet 5?
Its listed token prices are the same as Sonnet 5, but Anthropic says Sonnet 5.5 typically uses fewer tokens and can therefore cost up to 30% less per completed task.
Is Claude Sonnet 5.5 better for coding?
Anthropic reports substantial coding improvements over Sonnet 5, including higher scores on Terminal-Bench, FrontierCode and CursorBench. Actual results can vary by coding task and setup.
How fast is Claude Sonnet 5.5?
Anthropic says Sonnet 5.5 generates output more than 30% faster than Sonnet 5.
Is Claude Sonnet 5.5 available now?
Yes. Anthropic says Sonnet 5.5 is available across its supported platforms, including AWS, Google Cloud and Microsoft Azure, as well as the Claude Platform.
What is the difference between Sonnet 5.5 and Opus 5.5?
Sonnet 5.5 is designed for faster, well-scoped tasks and lower cost per task, while Opus 5.5 is aimed at more complex work requiring sustained judgment.
