Anthropic has launched Claude Sonnet 5.5, the second model in its Claude 5.5 series following the recent release of Opus 5.5. The company says the new model can generate responses more than 30% faster than its predecessor while reducing the overall cost of completing tasks by up to 30%.
The lower task costs do not come from a reduction in the price of individual tokens. Instead, Anthropic says Claude Sonnet 5.5 is designed to complete work using fewer tokens and fewer tool calls. The approach is intended to make the model more efficient for users who rely on AI for coding, research and other multi-step tasks.
Introducing Claude Sonnet 5.5, the second model in the Claude 5.5 family. It’s a clear upgrade over Sonnet 5, runs more than 30% faster, and costs up to 30% less for most work. https://t.co/UvXD8mDTF1
— Claude (@claudeai) September 28, 2026
Claude Sonnet 5.5 focuses on practical knowledge work
Anthropic has positioned Claude Sonnet 5.5 as a model for a wide range of everyday professional tasks. These include software development, debugging, document creation, presentation building and spreadsheet work. The company also highlights improvements in the model’s ability to produce polished outputs with attention to visual design and structure.
Coding is one of the areas where Anthropic reports the largest performance improvement. On the Terminal-Bench 4.0 coding benchmark, Claude Sonnet 5.5 scored 70.6%, compared with 10.3% for Sonnet 5. The company also reported a substantial improvement in chart-reading capabilities, with performance rising from 15.6% to 61.6%.
Anthropic said Sonnet 5.5 also comes close to Opus 5.5 on the GDPval-AA benchmark, which measures performance on professional knowledge-work tasks. However, the company continues to position Opus 5.5 as the stronger option for complex and open-ended work that requires extended reasoning and sustained judgement.
Early users have also reported faster speeds. Software company Box said some of its runs were completed 2.4 times faster, while Zendesk reported that its systems processed tickets 20% faster with the newer model. These results come from the companies’ own testing and do not necessarily represent performance across all workloads.
The release follows Anthropic’s broader push to make its Claude models more capable while improving their efficiency. By reducing the steps required to complete a task, the company aims to make Sonnet 5.5 useful for workloads where speed and operating costs both matter.
Lower costs come from efficiency rather than cheaper tokens
Despite the reduction in reported task costs, Anthropic has not changed Claude Sonnet’s API token pricing. Sonnet 5.5 costs US$2 per million input tokens and US$10 per million output tokens. Those rates match the previous Sonnet model and, according to the source material, match OpenAI’s GPT-6 Sol pricing.
The difference comes from how much work the model needs to perform to reach an answer. Anthropic says Sonnet 5.5 can use fewer tokens and fewer tool interactions, allowing users to complete some tasks without consuming as many resources. Lovable, a software development platform, reported that the model used about one-third fewer tool calls during its coding tests.
Anthropic also claims that lower-effort settings can outperform the previous Sonnet model’s highest results on some benchmarks while costing roughly one-tenth as much per task. The company says Claude applications will use Medium effort as the default setting.
View these figures as vendor-reported results rather than independent measurements. Benchmark scores can vary depending on testing methods, prompts and configurations, while the cost of an individual real-world task can depend heavily on its complexity and the tools involved.
For businesses using AI at scale, however, reducing tokens and tool calls could meaningfully lower operating costs. This is particularly relevant for coding agents and other systems that repeatedly interact with external tools while completing multi-step assignments.
New safeguards and wider cloud availability
Claude Sonnet 5.5 also introduces a new security feature within Anthropic’s Sonnet line. It is the first Sonnet model to include cybersecurity safeguards designed to manage higher-risk security requests. Anthropic says certain requests in this area will instead fall back to Sonnet 5.
The additional safeguards reflect AI models’ growing role in software development and cybersecurity. More capable coding models can assist with legitimate security work, but those same capabilities can also be used for potentially harmful activities. Anthropic’s approach is intended to distinguish higher-risk requests while continuing to support lower-risk uses.
Claude Sonnet 5.5 is available through major cloud platforms, including Amazon Web Services, Google Cloud and Microsoft Azure. Its availability across these services gives organisations using different cloud infrastructures a way to integrate the model into existing AI and software-development workflows.
Anthropic has also confirmed that Haiku 5.5 is scheduled to arrive in the coming weeks. The smaller model is expected to expand the company’s Claude 5.5 family beyond Sonnet and Opus. However, further details on its capabilities and pricing will depend on the company’s next announcement.
The release places Claude Sonnet 5.5 in an increasingly competitive market for AI models focused on coding and professional work. Anthropic is emphasising efficiency alongside benchmark performance, with the model designed to complete more work in fewer steps rather than simply reducing the price of individual tokens.
