Anthropic has launched Claude Sonnet 5.5, the second model in the 5.5 series following the flagship Opus 5.5. The company designed this version for daily tasks such as coding, document processing, and spreadsheet management, emphasizing higher speed and lower costs compared to its predecessor.
The model demonstrates significant progress in programming benchmarks. In Terminal-Bench 4.0, Sonnet 5.5 achieved a score of 70.6%, outperforming the 66.4% result of Opus 5.5 and the 10.3% score of the previous Sonnet 5. In CursorBench 4.0, the new model reached 55.5%, while Opus 5.5 scored 57.8%.
In the OSWorld 2.1 benchmark, Sonnet 5.5 scored 80.1%, closely trailing the 81.8% result of Opus 5.5. Additionally, the model is the first in the Sonnet line to complete the game Pokémon Red by analyzing screenshots, demonstrating its ability to process long sequences of visual information.
Sonnet 5.5 generates responses over 30% faster than Sonnet 5 and reduces the cost of typical tasks by up to 30% due to higher token efficiency. The pricing remains at $2 per million input tokens and $10 per million output tokens, with cache reads priced at $0.20 per million tokens.
The model is currently available via the Claude platform, API, and major cloud providers, including Amazon Web Services, Google Cloud, and Microsoft Azure. Anthropic notes that while Sonnet 5.5 is highly capable, Opus 5.5 remains the preferred choice for complex, open-ended tasks requiring deep reasoning.
