Anthropic has released Claude Opus 5.5, an upgrade to its Opus 5 model, with significant changes for API users. The new version offers a 20% reduction in per-token costs for both input and output, and a claimed 30% faster output speed. A notable change is the introduction of a caching mechanism, where writing a prefix costs more but subsequent reads are drastically cheaper, leading to substantial savings for repeated use cases. Developers are advised to re-evaluate their token usage and costs, as the efficiency gains can vary between 20% and 40% depending on the workload. AI
IMPACT API users will see reduced costs and faster response times, necessitating re-evaluation of token usage and pricing models.
RANK_REASON Frontier-lab model release with system card. [lever_c_demoted from frontier_release: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →