Anthropic launches Opus 5 AI model with fewer limits and lower cost

In short: Anthropic has released Opus 5, a new AI model it says will cost less and have fewer restrictions than its Fable 5 model.

What happened

Anthropic launched Opus 5 on Friday. The company describes it as the newest version of its “Opus” line, which it positions as a top tier model.

TechCrunch reports that Opus 5 is smaller than Anthropic’s Fable 5 model, but is expected to be cheaper and less restrictive for many common uses. Anthropic also says Opus 5 beats Fable 5 on several tests it shared in its announcement.

Anthropic said Opus 5 is “much stronger at verifying its work and iterating carefully until it succeeds.” In simple terms, it is meant to check its own answers more and keep trying when a task is hard, a bit like a student rechecking their homework before turning it in.

A key difference is data handling. Opus 5 is not covered by the 30 day data retention policy that applies to Fable and Mythos, according to the report. Data retention means how long a company keeps records of what users send, like keeping receipts after a purchase.

Anthropic says Opus 5 still has safety limits, especially around cybersecurity. For example, it will not help scan a software “binary” (a compiled app file) for weaknesses, but it may help review source code (the human readable instructions) to find issues for defensive fixes.

Anthropic is also introducing an opt-in beta feature called Automatic Fallbacks for API users (people connecting to the model through software). If a request triggers a safety check, the system can send it to a less powerful model so users get an answer instead of an error.

Why it matters

For businesses and everyday users, a cheaper model with fewer blocks can make AI tools easier to use day to day. At the same time, the remaining safety rules show Anthropic is still trying to limit harmful uses.

Source: TechCrunch AI