Anthropic launches Claude Opus 5 at half Fable 5's price

Anthropic's Opus 5 launch is more interesting for its price line than for the benchmark chart it comes with. The company shipped the model on July 24 at $5 per million input tokens, the same rate as Opus 4.8, and pitched it as reaching near-Fable-5 performance at half the price, Anthropic wrote.

The numbers behind that pitch are, for now, Anthropic's own. On Frontier-Bench v0.1 Opus 5 scored 43.3%, versus 18.7% for Opus 4.8 and 33.7% for Fable 5, per Decrypt. On Zapier's AutomationBench the pass rate runs about 1.5 times the next closest model at matching cost. On OSWorld 2.0, a computer-use benchmark, it clears Fable 5's peak score using roughly a third of the budget. Life sciences saw a jump of more than ten percentage points on organic chemistry evaluations over Opus 4.8.

The second structural change is a reasoning-effort toggle that lets a user pick low, medium, high or xhigh per prompt, trading cost against capability, Fortune reported. A Fast mode runs at roughly 2.5 times the default speed for twice the base price. The model ships with a 1M-token context window and becomes the default on Claude Max and the strongest option on Claude Pro.

The honest caveat is that these are launch-day benchmarks from the vendor, so real-world workload performance is still to be tested. The reporting also flags that Opus 5 sits behind Mythos 5 on cybersecurity tasks, which matters for security-heavy use cases. What the coverage does not give you is Fable 5's exact list price broken out next to Opus 5, or the methodology behind Frontier-Bench v0.1, so treat 'half the price of Fable' as company framing.

If the numbers hold in production, the winners are teams currently paying frontier prices for coding assistance and workflow automation. A same-cost, higher-capability step at the Opus tier is a rare thing in this release cycle, and it puts direct pricing pressure on whoever ships the next frontier-class model.