Claude Opus 5 Arrives: Near-Fable Performance at Half the Price

Anthropic's latest flagship, Claude Opus 5, landed on July 24, 2026 — and it does something unusual: it outperforms Anthropic's own top-of-the-range Fable 5 on the majority of benchmarks, while costing about half as much.

The pricing is unchanged from Opus 4.8 — $5 per million input tokens and $25 per million output tokens. That's exactly half of Fable 5's input price. Given that Opus 5 beats Fable 5 on 8 out of 13 benchmarks in Anthropic's own head-to-head comparison, the value-per-dollar argument is hard to dismiss.

What changed under the hood

The headline numbers are striking. On Frontier-Bench v0.1, Opus 5 more than doubles Opus 4.8's score and surpasses every other model tested. On ARC-AGI 3 — one of the most demanding reasoning benchmarks in existence — Opus 5 scores three times higher than the next-best competitor.

The model ships with a 1-million-token context window, 128K maximum output, and a May 2026 knowledge cutoff. One practical addition is a new low/medium/high effort toggle that lets developers trade compute cost against capability on a per-request basis. A quick summarization job doesn't need the same horsepower as a multi-step research synthesis, and now you can say so explicitly.

Where it sits in the lineup

Opus 5 is now the default model on Claude Max and the strongest available on Claude Pro. Fable 5 remains the absolute ceiling — Anthropic's most experimental and expensive tier — but for most professional workloads, Opus 5 appears to close that gap significantly.

Anthropic is also highlighting alignment improvements. According to TechCrunch, the company says Opus 5 is its most aligned model to date, with the lowest measured rates of deceptive behavior across its model family. That's increasingly relevant as enterprises weigh whether to trust AI systems with autonomous, multi-step tasks where the cost of subtle misbehavior compounds.

The bigger picture

This is Anthropic's fourth model release in roughly two months — a pace that was unthinkable for frontier labs even a year ago. The pattern fits what analysts have been calling the "compressed frontier": the gap between a $5-input model and a $10-input model is shrinking faster than anyone predicted, and the practical ceiling is moving upward with it.

For developers currently on Opus 4.8, the upgrade path is straightforward — same price, dramatically better performance. For those holding off on Fable 5 due to cost, Opus 5 makes the calculus much easier.

Claude Opus 5 is available now via the Anthropic API, Claude.ai, and through Amazon Bedrock and Google Cloud Vertex AI.