Claude Opus 5 vs Fable 5: Benchmarks, Price, Which to Use

Claude Opus 5 vs Fable 5: The $5 Model Just Beat the $10 One

Opus 5 costs half what Fable 5 does and wins most of Anthropic's launch benchmarks. But Fable still takes four of them.

Jul 25, 2026

On July 24, Anthropic released Claude Opus 5, and its benchmark scores put it ahead of Claude Fable 5, the more expensive flagship, on most of the tests published at launch.

On Frontier-Bench v0.1, a test of real software engineering tasks, Opus 5 hit 43.3% at max effort. Fable 5, the model Anthropic has spent weeks calling its most capable, scored 33.7%. For its pricing, Opus 5 costs $5 per million input tokens and $25 per million output tokens. Fable 5 costs $10 and $50.

That's awkward for Anthropic. Fable 5 was pulled globally within days of its June launch over a U.S. export control order, returned weeks later, and lost its place in standard subscription plans. Now its cheaper sibling beats it on the numbers that were supposed to justify all of that.

Claude Opus 5 vs Fable 5 Benchmarks: Who Actually Wins

Frontier-Bench isn't the only board where Opus 5 pulls ahead. It also beats OpenAI's GPT-5.6 Sol on that same test, 43.3% to 34.4%. Opus 5 loses four rows on Anthropic's table, though, and they're the ones worth knowing before you switch anything.