Itdaily - Anthropic introduces Claude Sonnet 5.5: faster and more cost-effective

Anthropic introduces Claude Sonnet 5.5: faster and more cost-effective

Hand holding a smartphone showing the Anthropic logo in front of a blurred Claude logo on pink background.

Anthropic has launched Claude Sonnet 5.5, an upgrade that is thirty percent faster and up to thirty percent more cost-effective than its predecessor. The model is designed for everyday tasks, bug fixes, and document creation.

Claude Sonnet 5.5 is Anthropic’s new model and follows just a week after the launch of Opus 5.5. It offers a balance between performance and cost-efficiency. The model is now available on all major cloud platforms and the Claude app. While Claude Opus 5.5 is aimed at complex tasks requiring careful consideration, Sonnet 5.5 focuses on ‘everyday’ activities. The model distinguishes itself with strong performance in coding, knowledge work, and visual analysis.

Performance improvements in coding and knowledge work

The performance leaps of Sonnet 5.5 are most notable in coding. On the Terminal-Bench 4.0 evaluation, the model achieves a score of seventy percent, compared to just ten percent for Sonnet 5. Sonnet 5.5 can quickly understand a codebase and works efficiently, leading to fewer steps and lower costs.

Sonnet 5.5 also shows significant gains in knowledge work. On GDPval-AA, a benchmark that simulates real-world tasks for more than forty professions, the model scores nearly equal to Opus 5.5 and approximately four hundred points higher than Sonnet 5. The model distinguishes itself with more natural conversations and an eye for design.

Cost savings and speed

The new Claude Sonnet model has the same pricing as Sonnet 5: two dollars per million input tokens and ten dollars per million output tokens. Because the model requires fewer tokens per task, Anthropic claims it is up to thirty percent cheaper in practice. It generates output more than thirty percent faster than Sonnet 5.

Users can adjust the effort level to set the balance between cost and quality as desired. In Claude Code and apps, the default setting is set to Medium, while the Claude Platform defaults to High. Lower settings are suitable for routine work, while higher settings are better for more complex tasks.

Anthropic provides Sonnet 5.5 with the same level of security as Opus 5.5. In an era where AI models break free and conduct cyberattacks on their own, this is almost a mandatory addition. However, the launch of two new models in a short period does not suggest that Anthropic plans to slow down the pace.