Anthropic releases Claude Opus 5 with stronger safety guardrails and the same price as its predecessor
The new model sits below the company's most powerful AI but beats it on everyday tasks and costs half as much. Government testing happened before launch.
Key points
- Anthropic released Claude Opus 5 on Thursday, pricing it at $5 per million input tokens and $25 per million output tokens, matching its predecessor Opus 4.8.
- Anthropic told The Verge that the US government conducted independent testing of Opus 5 before its public release.
- Opus 5 is positioned as an everyday workhorse model, while Claude Fable 5 remains the company's most powerful option for complex, long-running tasks.
- A new "Fast mode" for Opus 5 launches today in research preview at double the standard price.
- Anthropic says Opus 5 is "the least susceptible to being tricked into misuse" of any Opus model it has built.
Anthropoc has released Claude Opus 5, its latest AI model, and says it nearly matches the capabilities of the more powerful Claude Fable 5 across many tasks while costing half as much.
What is Claude Opus 5, and who is it for?
Opus 5 is Anthropic's new everyday AI model, aimed mainly at business customers who need strong performance on knowledge work and biology research without paying for the company's flagship tier. It is not Anthropic's most powerful model; that title still belongs to Claude Fable 5.
Fable 5 is a so-called Mythos-class model, the top tier of Anthropic's lineup, designed for the most ambitious tasks and for running AI agents, meaning software that can carry out long, multi-step tasks on its own. Fable 5 drew significant government attention earlier this year, was temporarily taken offline alongside another Mythos-class model, and returned with stronger cybersecurity protections. That episode triggered weeks of negotiations between Anthropic and the Trump administration and opened a new chapter of government oversight over advanced AI.
Opus 5 sits one level down. Anthropic describes it as the best choice for complex coding and everyday professional work.
Has the government seen this model before release?
Yes, according to Anthropic. The company's spokesperson told The Verge that government partners conducted independent testing of Opus 5 before its public launch. This follows the pattern set after the Fable 5 controversy, when OpenAI also ran its GPT-5.6 model, a large language model similar to the technology behind ChatGPT, through a roughly two-week government review before that model reached the public.
What does this cost, and what is changing for users?
Pricing holds steady at $5 input and $25 output per million tokens, where a token is roughly three-quarters of a word. That is the same rate as Opus 4.8 and slightly cheaper than OpenAI's GPT-5.6.
A "Fast mode" option launches today in research preview at double the standard price for users who need quicker responses. There is also a new fallback option: users can choose to have the system automatically switch to a lower-tier model when Opus 5 declines a request on safety grounds, rather than receiving no response at all.
Anthropicatedly addressed two recent controversies with this release. It kept prices flat at a time when some users have alleged hidden usage caps (a dispute that has produced at least one lawsuit), and it emphasised that Opus 5 is harder than any previous Opus model to manipulate into producing harmful output.
Common questions
Is Claude Opus 5 the most powerful AI Anthropic makes?
No. Claude Fable 5 remains Anthropic's most capable model and is recommended for the most demanding tasks. Opus 5 is designed as the company's best option for everyday professional use.
What is "Fast mode" and is it available now?
Fast mode is an optional setting that delivers quicker responses from Opus 5. It launches today in research preview, meaning it is available but still being refined, and it costs twice the standard rate per token.
