Anthropic releases Claude Opus 5, approaching Fable 5 capabilities

Anthropic released Claude Opus 5 on Thursday, days after an OpenAI security incident and weeks after its own regulatory friction with the U.S. government. The company described Opus 5 as a model that "comes close to the capabilities of Claude Fable 5 in many domains" while delivering significant improvements in complex coding tasks. This positioning reflects a deliberate segmentation: Fable 5, the public-facing Mythos-class model, remains Anthropic's flagship for the most ambitious types of work and long-horizon AI agent projects, while Opus 5 targets everyday knowledge work and biology applications for enterprise customers.

The regulatory context looms large. Fable 5 attracted government scrutiny and was taken offline for several weeks alongside Mythos 5 before being restored with stronger cyber safeguards. Those negotiations between Anthropic and the Trump administration marked the beginning of a new regulatory era for AI. When OpenAI subsequently released GPT-5.6, it entered a roughly two-week trial period limited exclusively to government-approved entities. Anthropic has confirmed that Opus 5 has likewise undergone independent government testing; spokesperson Danielle Ghiglieri told The Verge, "We continue to work with our government partners to conduct their own independent testing of our models. This includes Opus 5."

Anthropically's pricing and feature strategy address both market competition and recent customer complaints. Opus 5 is priced at $5 input and $25 output per million tokens—identical to its predecessor Opus 4.8—but the company emphasizes that it is "half the price" of Fable 5 and "slightly cheaper than OpenAI's GPT-5.6." The company also markets Opus 5 as having stronger cybersecurity safeguards than Opus 4.8 and as "the most aligned Opus model and is the least susceptible to being tricked into misuse," directly addressing the vulnerabilities that drew regulatory attention to Fable 5. To enhance performance flexibility, Anthropic is launching a "Fast mode" feature in research preview today, offering higher speeds for double the standard price. Additionally, the company is introducing an option for automatic fallbacks to a lower-tier model when its safeguards require declining a user request—a feature that appears to respond to recent controversies over hidden usage caps and concerns that AI labs are passing costs onto customers in opaque ways.

Opus 5's release occurs in a shifting regulatory environment. Fable 5, Anthropic's most capable model, drew government scrutiny that led to it being taken offline for a few weeks alongside Mythos 5 before returning with stronger cyber safeguards. The ensuing negotiations between Anthropic and the Trump administration marked the opening of a new era of AI regulation—one where major model releases now undergo government review. When OpenAI released GPT-5.6 shortly after, it underwent a roughly two-week trial period limited to government-approved entities, suggesting this oversight may become standard practice.

Anthropically's positioning of Opus 5 reflects both competitive and regulatory pressures. By pricing it identically to Opus 4.8 ($5 input / $25 output per million tokens) while claiming it is "half the price" of Fable 5, the company is defending against both cost-of-service complaints and competition with OpenAI's GPT-5.6. The emphasis on cybersecurity safeguards and being "the most aligned Opus model and is the least susceptible to being tricked into misuse" directly addresses the concerns that triggered Fable 5's regulatory attention. The addition of features like Fast mode and automatic fallbacks to lower-tier models when safeguards decline a request appears designed to address recent controversies over hidden usage caps and how AI costs are being passed to enterprise customers.

No comments yet. Be the first to share your thoughts!