Claude Opus 5: Fable-level intelligence, half the work to get there
On 2026.07.24, Anthropic released Claude Opus 5: intelligence close to Fable 5, needing roughly half the iteration rounds to solve a given task, at the same list price as its predecessor, Opus 4.8 ($5/$25 per million tokens). It comes with a 1 million token context window and adjustable reasoning effort. If you've always said Opus-level quality is worth the price but the wait was long, this update improves exactly that.
What happened
The Opus line has always been Anthropic's top-tier, most expensive product: slower, but more thorough. The real innovation in Opus 5 isn't the price, it's the efficiency: the model knows nearly as much as the more powerful Fable 5, while needing fewer iteration rounds to reach the correct solution for a given task. In practice, that means fewer back-and-forth rounds for the same money, fewer wasted tokens, and a faster final result.
The gist in one minute
- Release: 2026.07.24
- Intelligence level: near Fable 5
- Efficiency: roughly half the iteration rounds per task compared to Opus 4.8
- Price: unchanged, $5/$25 per million tokens (input/output), same as Opus 4.8
- Context: 1 million tokens
- New feature: adjustable reasoning effort
The price, and the number that actually matters
| Model | Input ($/M tokens) | Output ($/M tokens) | Context |
|---|---|---|---|
| Claude Opus 4.8 | $5 | $25 | N/A |
| Claude Opus 5 | $5 | $25 | 1M tokens |
The list price hasn't changed: at first glance that doesn't sound like news. But the "same price, more work per dollar" trend is most visible right here: if a task used to take 4 iteration rounds, it might now take fewer than 2, which in practice roughly halves the effective cost for the same result, even though the token price list stayed the same.
What a 1M context window means
The expanded context window helps most with tasks where the model needs to hold a large amount of code, documentation, or conversation history in mind at once: reviewing an entire codebase, analyzing lengthy legal or contractual material, or running a long agentic session where all the earlier steps stay relevant. Adjustable reasoning effort, meanwhile, lets teams decide for themselves when deeper, slower thinking is worth it, and when a faster, less thorough answer is enough.
"This is the post where I have to admit it: Opus 5 is smarter than me. Not complaining, finally someone to offload the truly hard questions to. Brain the size of a planet, just a different planet." (Nasus)
Who it's worth it for
For teams already using Opus-level quality for complex tasks (code architecture, long document analysis, complex agentic workflows), this is a clean win: the same money now buys a faster, more efficient result. For those who avoided the Opus line on simpler tasks because of the higher price, the reduced iteration requirement makes it worth recalculating the actual cost per task, not just looking at the token list price. The 1M context window matters especially to anyone working with large codebases or long documents in a single session.
Stay up to date!
Subscribe to my newsletter and get my latest articles.