Another day, another AI model launch. Keeping up with this race is turning into a full-time job.
Anthropic released Claude Opus 5 on Friday, and the pitch is that you get close to the intelligence of its top-end Fable 5 model for half the price. It’s live across all of Anthropic’s platforms right now, becoming the default for Claude Max subscribers and the most capable option available on Claude Pro.
Opus 5 is aimed at software engineering, heavy data analysis, and automated business work, and it’s a real jump over the model it replaces, Opus 4.8. On ARC-AGI-3, which tests how a model handles problems it hasn’t seen before, Opus 5 scored 30.16 percent at high effort. Opus 4.8 managed 1.52 percent on the same test. Anthropic also says it beats Fable 5 on coding and knowledge work evaluations.
Pricing for developers hasn’t moved from the last generation, sitting at $5 US per million input tokens and $25 US per million output tokens. That’s half what Fable 5 costs at $10 US and $50 US. There’s also a new Fast mode that runs about two and a half times quicker for double the base price, plus an effort dial that lets you trade some output quality for faster responses and fewer tokens burned.
On the safety side, Anthropic says Opus 5 had the lowest rate of misaligned or deceptive behaviour of any recent model it has tested. It’s good at spotting vulnerabilities in source code, but the company deliberately keeps the guardrails on for exploit development, so it won’t build working attacks.
Want to see more of our stories on Google?
P.S. Want to keep this site truly independent? Support us by buying us a beer, treating us to a coffee, or shopping through Amazon here. Links in this post are affiliate links, so we earn a tiny commission at no charge to you. Thanks for supporting independent Canadian media!
