Anthropic has launched Claude Sonnet 5.5, an updated model designed for bug fixing, documentation, presentations, and spreadsheet creation. The new version achieves 70.6% on Terminal-Bench 4.0, a benchmark that measures programming task execution directly in terminals—a 585% improvement over Sonnet 5. Notably, it became the first model in the Sonnet line to complete Pokémon Red using only screenshots.
Despite maintaining the same API pricing as its predecessor, Sonnet 5.5 requires up to 30% fewer tokens to accomplish the same tasks and runs 30% faster. This efficiency gain makes it a more cost-effective option for developers and teams relying on token-based billing, delivering better performance without price increases.