Anthropic releases Claude Opus 4.7, a less risky model than Mythos

Anthropic on Thursday introduced a new artificial intelligence fashions, Claude Opus 4.7which the corporate stated is an enchancment over previous fashions however is “less broadly capable” than its most up-to-date providing, Claude Mythos Preview.

Claude Opus 4.7 is healthier at software program engineering, following directions, finishing real-world work and is its strongest usually accessible model, Anthropic stated. But the model’s cyber capabilities will not be as superior as Claude Mythos Preview, which Anthropic rolled out to a choose group of corporations as a part of a new cybersecurity initiative known as Project Glasswing earlier this month.

“We are releasing Opus 4.7 with safeguards that automatically detect and block requests that indicate prohibited or high-risk cybersecurity uses,” Anthropic stated in a launch. “What we learn from the real-world implementation of these safeguards will help us work towards our eventual goal of a broad release of Mythos-class models.”

Since its founding in 2021, Anthropic has spent years rigorously crafting its fame as a agency that is extra devoted to secure and accountable AI deployment than rivals like OpenAI.

The launch of Project Glasswing has sparked a variety of high-profile conferences amongst members of the Trump administrationtech CEOs and financial institution chief executives about the security risks of highly effective AI fashions.

Anthropic doesn’t plan to make Claude Mythos Preview usually accessible, however the firm has stated its objective is to study the way it may ultimately deploy Mythos-class fashions at scale.

The launch of Claude Opus 4.7 on Thursday comes after Anthropic launched Claude Opus 4.6 in February. Anthropic stated the brand new model outperforms Claude Opus 4.6 throughout many use circumstances, together with trade benchmarks for agentic coding, multidisciplinary reasoning, scaled instrument use and agentic laptop use, in accordance with to release.

Anthropic stated it experimented with efforts to “differentially reduce” Claude Opus 4.7’s cyber capabilities throughout coaching.

The firm inspired safety professionals who’re keen on utilizing the model for “legitimate cybersecurity purposes” to use by a formal verification program.

Claude Opus 4.7 is offered throughout all of Anthropic’s Claude merchandise, its utility programming interface and thru cloud suppliers Microsoft, Google and amazon. The new model is similar value as Claude Opus 4.6, Anthropic stated.

WATCH: Anthropic co-founder says company in process of broadening Mythos access: Semafor event