Anthropic released Claude Opus 5 on Thursday, positioning it as the new default model for most users and pitching it as nearly matching the capability of its top-tier Fable 5 model at half the price.
It’s the fourth model the company has shipped in under two months, following Mythos 5, Fable 5, and Sonnet 5 in June — an unusually rapid release cadence even by AI industry standards.
What’s actually new
According to Anthropic’s own announcement, Opus 5 sets a new internal state-of-the-art on coding and knowledge-work evaluations, including benchmarks the company calls Frontier-Bench and GDPval-AA, while remaining behind Mythos 5 specifically on cybersecurity tasks.
Pricing stays flat at $5 per million input tokens and $25 per million output tokens, unchanged from predecessor Opus 4.8, meaning users get a capability jump without a price increase.
The model becomes the new default on Claude Max, Anthropic’s premium consumer tier, and is now the strongest model available to Claude Pro subscribers.
One practical addition: Opus 5 includes a toggle letting users set how much computational “effort,” low, medium, or high, the model spends on a given task, aimed squarely at enterprise customers who’ve grown wary of unpredictable AI bills.
Anthropic also says the model requires less back-and-forth to get right, verifying its own work and recovering from errors without the user needing to step in.
Positioned for science and safety-conscious users
Anthropic is marketing Opus 5 partly toward the scientific research community, calling it its most capable generally available model for that use case and highlighting notably stronger biology-related performance compared to its predecessor.
On the safety side, the company describes Opus 5 as its most aligned Opus model to date and the hardest in the lineup to manipulate into misuse — a message clearly aimed at enterprise and institutional buyers who need assurances before deploying AI at scale.
Unlike Fable 5, Opus 5 does not retain user data for 30 days, according to the company.
The bigger context: a very crowded two months
To understand why Opus 5 matters, it helps to see where it sits in Anthropic’s recent run of releases.
In April, Anthropic gave a small group of users early access to a research model called Claude Mythos Preview, notable for unusually strong cybersecurity capabilities, including the ability to find and exploit software vulnerabilities.
That work fed into two June releases: Mythos 5, which remains restricted to a limited set of partner organizations because of its hacking-relevant capabilities, and Fable 5, a general-release sibling model built with additional safeguards.
Fable 5 briefly became a flashpoint after the US government imposed a temporary export-control-related suspension on it and Mythos 5 days after launch, reportedly tied to national-security concerns; Anthropic pulled the model, then restored access after roughly two weeks of negotiation with regulators. Sonnet 5 followed later in June as a mid-tier release.
Opus 5 arrives against that backdrop as the more cost-conscious, everyday-use answer — explicitly framed by Anthropic as approaching Fable 5’s ability without Fable 5’s price tag or its more restrictive access requirements.
The company still recommends Fable 5 for the most demanding, long-running autonomous projects.
2026-07-24 23:34:00
