Anthropic has released Sonnet 5.5, a mid-tier model designed to operate faster and at a lower cost than the previous version. The company states the new system is thirty percent quicker than Sonnet 5 while reducing token consumption. This update targets everyday tasks such as coding and document creation, positioning the model as a practical assistant for routine workflows.
Earlier this year, the firm highlighted Sonnet 5 for its efficient agentic deployment, allowing lower-cost agent execution compared to rivals. The current release shifts focus toward speed and specific performance metrics. Benchmarks indicate Sonnet 5.5 outperforms the more powerful Opus 5.5 on agentic coding tasks, particularly when spawning multiple agents without hitting cost ceilings. Additionally, the model now carries cyber safeguards previously reserved for the Fable and Opus tiers, matching their security standards. This move follows recent announcements from OpenAI and Meta, which have also introduced new models for their respective ecosystems.
* Sonnet 5.5 reduces token burn significantly compared to its predecessor
* The model applies the same cyber safeguards as the Opus and Fable tiers
* A new version of the Haiku model is expected in the coming weeks
