Claude Opus 5 launches with stronger cyber safeguards
Anthropic says Claude Opus 5 nears Fable 5 in some areas while adding stronger misuse protections after government scrutiny.
By Hana Yoshida · Markets Reporter
3 min read
Anthropic released Claude Opus 5 on Thursday, introducing a new AI model that the company says approaches Claude Fable 5 in several areas while adding stronger cybersecurity controls. The launch matters because Anthropic is positioning Opus 5 as a safer, cheaper everyday model for business users after recent government concern over its more powerful systems.
Anthropic said Opus 5 is much improved at complex coding work and is aimed heavily at enterprise customers. The company described it as suited for knowledge work and biology, while saying Claude Fable 5 remains the better choice for the most ambitious tasks and long-running AI agent projects.
What is Claude Opus 5?
Claude Opus 5 is Anthropic’s newest model in its Opus line, designed as a general-purpose option for demanding work. Anthropic is presenting it as close to Fable 5 in many domains, but with a lower price and a stronger safety posture than the previous Opus model.
The release follows scrutiny of Claude Fable 5, a public-facing Mythos-class model. According to The Verge, Fable 5 and Mythos 5 were taken offline for several weeks after government concerns, then returned with tighter cyber safeguards.
That episode led to weeks of talks between Anthropic and the US government, The Verge reported. The publication also reported that OpenAI’s GPT-5.6 later had a roughly two-week limited preview for government-approved entities before wider availability.
Did the government test Claude Opus 5?
Anthropic spokesperson Danielle Ghiglieri told The Verge that the company continues working with government partners so they can conduct independent testing of its models. Ghiglieri said that work includes Opus 5.
Anthropic said Opus 5 has stronger cybersecurity protections than Opus 4.8. The company also said the new model is its most aligned Opus release and is the least prone among Opus models to being manipulated into misuse.
The added safeguards are a central part of Anthropic’s pitch. They arrive days after an OpenAI security incident became a major topic in the tech industry, according to The Verge.
How much does Claude Opus 5 cost?
Anthropic is keeping Opus 5 pricing the same as Opus 4.8: $5 per million input tokens and $25 per million output tokens. The company said that makes Opus 5 half the price of Fable 5, and The Verge noted that it is slightly cheaper than OpenAI’s GPT-5.6.
The pricing details arrive as Anthropic faces questions about usage limits and broader concerns that AI lab costs are being passed on to customers in new ways. The Verge cited at least one lawsuit tied to allegedly hidden usage caps.
Anthropic is also adding a “Fast mode” for Opus 5 in research preview. Customers who choose it get higher speeds at double the regular price.
The company will also let customers opt in to automatic fallback to a lower-tier model when Opus 5’s safeguards require it to decline a request. That option gives users a way to continue a task with a different model when Opus 5 refuses to answer.
This story draws on original reporting from The Verge.