Anthropic's latest Claude model boosts speed and cuts costs
Anthropic has released Claude Opus 5.5, claiming it matches the performance of its larger Fable 5.1 model for most tasks while reducing operational costs by about 40%. The company also notes a 30% increase in output speed and a 20% reduction in token prices, along with higher usage limits for subscribers. The model is available immediately, with smaller variants expected in the coming weeks.
The release arrives less than two months after Opus 5, which itself promised near-Fable performance at half price. Anthropic reports output generation over 30% faster, with token prices down 20%. Box's VP of AI Products noted the model used a third of Opus 5's tokens while producing answers 40% less verbose without losing accuracy.
Subscribers gain a 20% increase in five-hour usage limits, and slower token burn may boost effective run capacity by roughly half. Sonnet 5.5 and Haiku 5.5 arrive in coming weeks. The announcement also stresses safety, citing CEO Dario Amodei's writing on moderating AI capability advances.
This release could shift how businesses and individuals deploy frontier AI. Lower token costs and faster output may encourage broader automation in document-heavy sectors like finance and government. Higher usage limits could reduce subscription friction, normalizing heavier daily reliance on AI assistants. The safety emphasis suggests Anthropic is managing risk expectations. If performance claims hold, competitors may face pressure to match price and speed, potentially accelerating adoption while raising questions about job displacement and over-reliance on automated systems.