Anthropic released Claude Opus 5 on Thursday, pitching the new AI model as a high-end option that gets near Claude Fable 5 in many areas while adding tougher cybersecurity guardrails. The timing is not subtle: Anthropic has been dealing with US government concern over its more powerful models, and OpenAI recently had a security incident that put AI model security back in the industry’s bloodstream.
Anthropic said in its announcement that Opus 5 is “close” to Fable 5’s abilities across many domains and performs better on complex coding work than the prior Opus model. That is the company’s claim, not an independent benchmark, and Anthropic is clearly trying to place Opus 5 near the top of its lineup without making it the wildest thing in the rack.
What is Claude Opus 5?
Claude Opus 5 is Anthropic’s newest Claude model, aimed especially at enterprise customers. Anthropic describes it as suited for knowledge work and biology, and as an everyday model for organizations that want strong performance without using Fable 5 for everything.
Fable 5 remains Anthropic’s stated choice for its most ambitious work, including long-horizon AI agent projects. In plain English, those are systems meant to carry out multi-step tasks over extended periods rather than answer one prompt and go back to sleep. That distinction matters because agents can chain actions together, which is useful in a company workflow and a headache when safety testing is weak.
Why are cyber safeguards part of the launch?
The security framing follows earlier tension between Anthropic and the US government over Fable 5. Fable 5 is a public-facing Mythos-class model that drew government criticism, was taken offline for several weeks along with Mythos 5, and later returned with stronger cyber safeguards, according to prior reporting by The Verge.
Those concerns helped trigger weeks of talks between Anthropic and government officials and became part of a broader shift in AI regulation under the Trump administration, according to The Verge. OpenAI’s GPT-5.6 rollout also reflected that new posture: it had a roughly two-week limited preview that was restricted to entities approved by the government.
Asked whether Anthropic showed Opus 5 to the Trump administration before releasing it publicly, Anthropic spokesperson Danielle Ghiglieri told The Verge that the company continues working with government partners on independent model testing and said that work includes Opus 5.
That answer leaves some daylight. It confirms government-linked testing is part of the process, but it does not say the administration cleared the model before launch. For customers, the practical takeaway is narrower: Anthropic is selling Opus 5 as a safer, enterprise-friendly model that approaches Fable 5 in some areas, while reserving Fable 5 for the harder and longer-running AI jobs.
This story draws on original reporting from The Verge.
