Anthropic releases Claude Opus 5 at half the price of its flagship Fable 5

Claude Opus 5 is here, and it resets the price of frontier AI.

The release

On July 24, 2026, Anthropic released Claude Opus 5 at $5 per million input tokens and $25 per million output tokens. That is the same price as Opus 4.8, and half the price of the flagship Fable 5.

Holding the price flat while replacing the model is the quiet headline. The usual pattern in frontier AI is that each capability jump arrives with a premium attached. Opus 5 breaks it: the capability jumped, the invoice did not.

What the money buys

The benchmark picture, per Anthropic:

• More than double Opus 4.8’s performance on FrontierBench v0.1, at a lower cost per task
• Within 0.5 percent of Fable 5’s peak score on CursorBench 3.2, at half the cost
• An ARC-AGI 3 score three times as high as the next-best model

On availability: Opus 5 is the new default model in Claude Max and is live in Claude Code and the API under the identifier claude-opus-5. For latency-sensitive work there is a fast mode running at around 2.5 times the speed for twice the base price, a straightforward trade of dollars for wall-clock time.

Why the pricing is the story

The CursorBench number is the one that changes budgets. Being within half a percent of the flagship at half the cost collapses the argument for defaulting to the most expensive model. For AI agents, coding assistants, and everyday knowledge work, the math shifts: near-flagship intelligence no longer requires flagship pricing.

That has compounding consequences for anything agentic. Agent workloads burn tokens in volume, so a 2x cost difference at near-parity capability is not a rounding error, it is the difference between a pilot and a production deployment. Teams that priced out agent systems on flagship rates now get to redo the spreadsheet with half the number in the cost column.

Would you switch your default model for half the price? For a lot of builders, that question now answers itself.