July 1st, local time 30thAnthropic Launch a new generation of medium-sizedModel Claude Sonnet 5, core enhanced AI AgentCapacity. Officially, the model ' s performance in self-planning, browser and end-of-service tools has reached the level of a previously large and expensive model。
For performance, Sonnet 5 scored 63.2% in the programming baseline test, close to the top model Opus 4.8 and significantly superior to the previous Sonnet 4.6。
Early partner feedback indicated that the model was able to carry out its own multi-platform tasks of updating Salesforce accounts, dispatching bulletins, etc., and had enhanced self-censorship. In terms of safety, the new model has been upgraded in terms of refusal of malicious requests, reduction of hallucinogenic rates, and resistance to infusion。
Sonet 5 became the default model for free and Pro versions as of this date. On the price side, the initial price of the model in Claude Platform is $2 per million input Token and $10 per million output (concession to 31 August), which will be adjusted to $3 and $15。
Anthropic emphasizes that the goal of Sonnet 5 is to significantly reduce the cost of a developer running AI smarts while maintaining high performance。