Artificial intelligence company Anthropic on Friday, July 24, released Claude Opus 5, its newest model and one it says comes close to the intelligence of its top-of-the-range Claude Fable 5, but at about half the price.
The launch follows Opus 4.8, which arrived on May 28, and joins the wider Claude 5 family that already includes Mythos 5, Fable 5 and Sonnet 5. A Haiku version is still waiting for its upgrade to the new generation.
According to Anthropic, Opus 5 actually beats Fable 5 on several tests covering coding and knowledge work, scoring 43.3 percent on the Frontier Bench v0.1 benchmark against Fable 5’s 33.7 percent. The company says the model is good at checking its own work and carefully retrying until it succeeds, such as writing computer-vision code from incomplete instructions.
Pricing stays the same as its predecessor, at 5 dollars per million input tokens and 25 dollars per million output tokens. Opus 5 becomes the new default on the premium Claude Max tier and the strongest option on Claude Pro, and unlike Fable and Mythos it is not subject to a 30-day data-retention rule.
On safety, Anthropic says its automatic filters are expected to trigger 85 percent less often on Opus 5 than on Fable 5. The model still blocks certain cybersecurity tasks, such as hunting for weaknesses inside compiled software, though it is allowed to review source code. A new beta feature called Automatic Fallbacks quietly reroutes a request to a smaller model when a safety filter kicks in, so users get an answer instead of an error.
Why it matters: cheaper access to a near-flagship AI makes powerful tools more affordable for the developers, students and small businesses who cannot pay premium rates, including many across Africa.
Source: TechCrunch.
Comments 0