Claude Sonnet 5.5 Now Faster with No Price Increase

Post Views: 10

Claude Sonnet 5.5 shows improved performance with same pricing, enhanced speed, and optimized costs.

Performance Enhancements

Claude Sonnet 5.5 demonstrates improved performance metrics without adjustments to pricing structures. The latest iteration of the AI model, designed for software development and administrative tasks, exhibits a 30% enhancement in response speed compared to its predecessor. This advancement is accompanied by optimized token usage, enabling reduced operational costs while maintaining the same API pricing framework.

Accessibility and Configuration

The model is accessible through cloud infrastructure providers including Amazon Web Services, Google Cloud, and Microsoft Azure. Developers can access the updated version via the Claude Platform using the identifier claude-sonnet-5-5. Users operating the model in non-thinking mode must transition to the between_tools configuration prior to implementation.

Benchmarking Results

Benchmarking results reveal that Sonnet 5.5 achieved a 70.6% score on Terminal-Bench 4.0, a multi-step command-line interface evaluation, surpassing the 10.3% score of Sonnet 5. On GDPval-AA, a comprehensive assessment spanning 44 occupational roles and nine industries, the model attained 1,844 points—2 points lower than Opus 5.5 but 395 points higher than Sonnet 5.

Terminal-Bench 4.0 Evaluation

Terminal-Bench 4.0 evaluates the model’s capacity to execute complex professional workflows within a terminal environment.

GDPval-AA Assessment

The model attained 1,844 points on GDPval-AA, 2 points lower than Opus 5.5 but 395 points higher than Sonnet 5.

Cost Analysis

Cost analysis indicates that Sonnet 5.5 offers lower expenses per task than Opus 5.5 at lower effort settings, while achieving comparable benchmark results at higher settings with similar cost structures. In the FrontierCode coding evaluation, Sonnet 5.5 outperformed Sonnet 5 by 10 points under high-effort conditions, achieving this at approximately one-fifteenth the cost per task.

User Feedback and Adoption

Early adopters reported enhanced comprehension of codebases and streamlined tool integration, reducing the number of required steps. A representative from CodeRabbit noted that the model exhibits superior decision-making across varying complexity levels, eliminating the predecessor’s reliance on web searches and excessive token consumption.

A representative from CodeRabbit noted that the model exhibits superior decision-making across varying complexity levels, eliminating the predecessor’s reliance on web searches and excessive token consumption.

Future Implementation

The organization plans to transition moderate and simple tasks to the new model in the coming weeks. Performance metrics on computer-use and chart-recognition assessments approached those of Opus 5.5. The model exceeded Sonnet 5 and GPT-6 Sol on AA-Briefcase, a test measuring long-term knowledge work capabilities.

Usability Improvements

Users described interactions as more intuitive, with the ability to follow slide templates for deck creation requiring minimal post-processing. Adjustments to effort settings allow users to balance cost, speed, and output quality, with default configurations set to Medium in Claude Code and Anthropic applications, and High on the Claude Platform.

Safety and Compliance

Safety evaluations involving 1,850 automated scenarios showed Sonnet 5.5 matched or exceeded Sonnet 5 in alignment, misuse resistance, and transparency. Anthropic confirmed no evidence of conflicting goal pursuit by the model. While testing cannot identify all potential failures, the system incorporates cybersecurity safeguards.

Operational Considerations

Routine software development tasks, including bug detection and resolution, remain viable. Higher-risk cybersecurity operations will default to Sonnet 5. The model retains the same biological safety protocols as its predecessor, which may occasionally flag legitimate microbiology and virology queries.

Access and Updates

Organizations seeking expanded access can apply to Anthropic’s verification programs. Additional developments in agentic AI, cybersecurity practices, and Anthropic’s Claude Code platform are highlighted in related coverage. Recent reports include mitigation of a zero-day vulnerability exploited in advanced cyberattacks, data breaches via SQL injection flaws, and security management solutions.