Anthropic has released Claude Sonnet 5.5, the second model in its 5.5 series, following the flagship Claude Opus 5.5. The new model is designed for daily tasks, programming, document processing, and presentation creation.
Performance metrics show a 30% increase in response speed and up to 30% lower costs for most tasks compared to the previous version. The pricing remains $2 per 1 million input tokens and $10 per 1 million output tokens, with cached tokens costing $0.20 per 1 million.
In programming benchmarks, Sonnet 5.5 scored 70.6% in Terminal-Bench 4.0 and 55.5% in CursorBench 4.0, significantly outperforming its predecessor. The model supports a context window of up to 1 million tokens and an output of 128,000 tokens.
The model demonstrates capabilities approaching the flagship Opus 5.5 in specific tests, such as the GDPval-AA v2.1 benchmark, where it scored 1844 points against 1846 for Opus 5.5. Opus 5.5 remains twice as expensive at $4 and $20 per million tokens.
Security features have been upgraded, including new protective mechanisms and anti-distillation classifiers to prevent data extraction. The model is currently available via the Claude chatbot, AWS, Google Cloud, Microsoft Azure, and API.
Anthropic plans to release the Claude Haiku 5.5 model in the coming weeks, targeting high-load scenarios and cost-sensitive applications.
