Anthropic has launched Opus 5, the latest version of its flagship AI model, continuing the rapid expansion of its 5-series lineup just two months after the release of Opus 4.8.
The company said Opus 5 is designed to deliver stronger performance at a lower cost than Fable 5 while introducing fewer restrictions for enterprise users.
Follow THE FUTURE on LinkedIn, Facebook, Instagram, X and Telegram
A Premium Model With A More Practical Proposition
Although Opus 5 is smaller than Fable 5, Anthropic said it is cheaper to run and less restrictive. The company also said the model outperformed Fable 5 on several internal benchmarks, suggesting improved efficiency without sacrificing performance.
Opus 5 follows the recent launches of Mythos 5, Fable 5 and Sonnet 5, leaving Haiku as the only model in Anthropic’s portfolio yet to receive a 5-series upgrade.
Anthropic Emphasizes Reliability And Iteration
According to Anthropic, Opus 5 is “much stronger at verifying its work and iterating carefully until it succeeds.” As an example, the company said the model independently built a computer vision pipeline from an incomplete prompt during internal testing.
The release reflects Anthropic’s continued focus on improving reasoning capabilities and reducing the amount of human intervention required for complex tasks.
Fewer Restrictions, But Cybersecurity Guardrails Remain
Unlike Fable 5 and Mythos, Opus 5 is not subject to Anthropic’s 30-day data retention policy. The company said the model is intended for customers seeking greater flexibility while maintaining existing security protections.
Restrictions remain for cybersecurity-related tasks. Anthropic said Opus 5 cannot be used to scan software binaries for vulnerabilities, although it can analyse source code for security flaws, which the company classifies as defensive security work.
Anthropic also said its safety classifiers are expected to activate 85% less frequently for Opus 5 than for Fable 5.
A Smaller Friction Point For Developers
The company is also rolling out a beta feature called Automatic Fallbacks. If a prompt triggers a safety classifier, participating API users will automatically be routed to a less capable model instead of receiving an error.
Anthropic said the feature is intended to reduce workflow interruptions for developers while maintaining existing safety measures.
