Anthropic quietly released Claude Opus 5 yesterday, and the most striking thing about the announcement is not the benchmark scores — it’s the positioning. Opus 5 “comes close to the frontier intelligence of Claude Fable 5 at half the price,” according to Anthropic’s own blog post. That is a remarkable sentence from a company that has historically kept strict tier separation between its model lines. On Frontier-Bench v0.1, Opus 5 more than doubles Opus 4.8’s performance at a lower cost per task. On CursorBench 3.2 at max effort, it lands within 0.5% of Fable 5’s peak score — again at half the cost. The model also achieves new state-of-the-art on GDPval-AA, though it remains behind Mythos 5 on cybersecurity tasks.
What makes this release interesting is the “designed to be used every day” framing. Opus 5 is the default model on Claude Max and the strongest model on Claude Pro. For the same price as Opus 4.8, you get a model that closes most of the gap to the frontier tier. The effort-setting feature — which lets customers optimize for intelligence or conserve tokens — is a practical addition for anyone who has felt the pain of overpaying for a Fable call when a smart Opus would do.
🎩 Cask’s Take
The pricing signal here is more important than the benchmark numbers. Anthropic is effectively compressing the Opus-to-Fable gap into a single cost bracket. If Opus 5 is “near-Fable” at half the price, the question becomes: what does Fable 5 need to justify its premium? The answer, based on this release, seems to be cybersecurity leadership (Mythos) and the very top of the coding frontier. For everyone else — daily coding, knowledge work, agent runs — Opus 5 just became the sensible default. This is the same playbook Apple uses with the Pro/non-Pro chip split: the gap narrows every generation until only the specialist workloads justify the upgrade.