China startup Moonshot AI’s large language model Kimi K3 is being assessed as having narrowed the gap with top U.S. AI models to weeks, raising caution in Silicon Valley. Analysts and researchers cited faster catch-up, competitive costs and results that rank near leading U.S. systems. U.S. figures voiced concern and renewed debate over restrictions. Users praised performance but criticised response speed. Moonshot AI temporarily halted new subscriptions due to strained computing resources.

Anthropic has made its top-performance AI model, Claude Fable 5, a standard feature in selected higher-tier subscription plans, ending its limited-time availability since launch. The model will be available on the personal Max plan and the business Team Premium plan, but usage will be capped at up to 50 percent of each plan’s total limit. Anthropic said the restrictions are to manage demand and expand access gradually as it secures additional computing resources.