Anthropic has unveiled Claude Sonnet 5.5, its latest artificial intelligence model, designed to enhance the execution of daily tasks while reducing costs. This new model is part of the Claude 5.5 family, alongside Claude Opus 5.5, which is geared towards more complex tasks requiring continuous reasoning and judgment. Sonnet 5.5 focuses on specific daily tasks that demand speed and efficiency.
The new model boasts several improvements over its predecessor, Claude Sonnet 5, launched three months prior. Key enhancements include a more than 30% increase in response speed, reduced token consumption, and significant advancements in programming and cognitive tasks. These improvements make Sonnet 5.5 suitable for tasks that require rapid iteration and adjustment, particularly when high-level reasoning is not necessary.
Anthropic highlights that Sonnet 5.5 achieves results at over 30% faster than Sonnet 5, making it the fastest model in the Sonnet series. The cost-effectiveness is not due to changes in token pricing but rather the model's ability to complete tasks using fewer tokens. Tests indicate that Sonnet 5.5 may cost around 30% less per task compared to its predecessor, despite unchanged token prices.
The pricing structure for Claude Sonnet 5.5 remains the same as Sonnet 5, with charges of $2 per million input tokens, $10 per million output tokens, and $0.20 per million tokens for reading cache. The model's efficiency in using fewer tokens for tasks translates to lower costs for users.
Claude Sonnet 5.5 demonstrates a significant leap in programming tests, particularly in the Terminal-Bench 4.0, where it scored 70.6%, compared to 10.3% for Sonnet 5. It also performed closely to Opus 5.5 in CursorBench, which simulates real-world coding tasks. The model has become faster at understanding software structures and more efficient in using programming tools.
Beyond programming, Sonnet 5.5 shows improvement in various cognitive tasks that mimic real-world job applications. In the GDPval-AA test, which evaluates models across 44 professions and nine industries, Sonnet 5.5's performance is close to Opus 5.5 and significantly better than Sonnet 5. The model also exhibits enhanced capabilities in computer usage, graph reading, and long-term task execution.
Anthropic aims to utilize Sonnet 5.5 in office work, including the creation of presentations, documents, and spreadsheets. Early testers have noted improvements in the model's ability to communicate and collaborate with users, with a greater emphasis on formatting and design. The model can follow templates for presentations and add visual enhancements to user interfaces.
Key points
- Claude Sonnet 5.5 offers over 30% faster response speed and is designed to reduce costs by using fewer tokens for tasks.
- The model demonstrates significant improvements in programming, cognitive tasks, and document creation.
- Anthropic plans to launch Claude Haiku 5.5, a smaller model for high-volume, low-cost task execution, in the coming weeks.
