Anthropic is continuing to release new AI models at a rapid clip, despite its CEO’s pleas for a unilateral slowdown to give the industry’s safety frameworks time to catch up with the technology’s capabilities. In this case, at least, we’re talking about the release of a new entry in the company’s mid-level Sonnet tier.
The company unveiled Sonnet 5.5 on Monday, describing it as a faster, cheaper alternative to Opus 5.5, which made its debut less than one week ago. “Where Opus 5.5 is built for complex work requiring careful judgment, Sonnet 5.5 is strongest at well-scoped everyday tasks, fixing bugs, and creating polished documents, slides, and spreadsheets,” Anthropic wrote in its announcement. Haiku 5.5, the third and final addition to the Claude 5.5 family of models, will be available in the coming weeks, the company added.
OpenAI also released two new models last week—GPT-6 Sol and Luna—both of which are similarly being sold as leaner and more cost-effective versions of the larger, more powerful Astra.
Anthropic’s Monday announcement touted the usual performance gains that are typical of a new model release from a frontier lab. Compared to its predecessor (Sonnet 5), Sonnet 5.5 scored higher on an evaluation testing agentic coding capabilities. Anthropic also noted that it’s “a better partner for collaboration” and that it generates outputs at least 30% quicker than Sonnet 5, making the new model “well suited to fast iteration on less complex tasks.”
Regarding safety and alignment, Anthropic said “Sonnet 5.5 doesn’t advance the frontier of our model’s capabilities,” meaning there were no new guardrails added. The announcement did note, however, that the model was the least likely of all the models Anthropic had tested internally to actively search for vulnerabilities in its testing sandbox.
Despite quickly winning support from the leaders of other leading AI labs, Anthropic CEO Dario Amodei’s call for the AI industry to “pace the frontier” earlier this month has thus far not produced any meaningful changes to the status quo in Silicon Valley. And even though the risks of runaway AI have been a major theme at the United Nations’ General Assembly since it began last week, Trump’s Washington summit with Chinese president Xi Jinping doesn’t seem to have made any meaningful headway on the subject of working together to mitigate those risks. According to the White House, the leader of the free world did, however, get his Chinese counterpart to agree to use his preferred phrase, “super intelligence.”
Read the full article here
