On 24 July, Anthropic shipped Claude Opus 5 - by the company's own numbers, nearly as capable as the flagship Fable 5, at half the price. The real news is the effort dial: you decide how hard the model thinks on each task, and how much you pay for it.
- On 24 July, Anthropic shipped Claude Opus 5: $5 per million input tokens and $25 per million output tokens - the same as the previous Opus 4.8.
- New is the effort dial (low/medium/high) and a fast mode - 2.5 times faster at double the price.
- Opus 5 becomes the default model for Claude Max, is API-accessible right away, and is available immediately in the app and in Claude Code.
I opened Anthropic's page and the first thing I saw wasn't the model. It was the price - the same as the previous one. The show isn't on stage here. The show is in the price list.
I'll set the benchmarks aside. Every vendor measures its own chip or model, which is why the line above says 'by the company's own figures'. What's interesting is the effort dial. Until now the model decided on its own how hard to think about your task. Now you decide. An easy task gets low effort and low cost; a hard one - you turn the dial up and pay more.
The direction's been building for months. The race is shifting from 'who's smartest' to 'who's smartest per euro'. Top-tier intelligence stays a luxury for the few genuinely hard tasks. Everyday work wants something else - a predictable, cheap, good-enough brain. That's why this launch is quiet. And why it matters.
Going forward, you're not paying for access to intelligence. You're paying for judgment - your judgment of how much you need right now. That's exactly where the craft is shifting to: knowing when a task needs the top tier and when medium is enough. Whoever learns that will do the same work for a fraction of the cost of someone who keeps the dial maxed out of vanity.