Anthropic released Claude Opus 5 on July 24, a model that not only costs exactly half as much as its flagship Fable 5 but also outperforms it on most major benchmarks. Priced identically to its predecessor Opus 4.8 at $5 per million input tokens and $25 per million output, Opus 5 immediately becomes the default on Claude Max and the strongest model on Claude Pro, effectively filling the consumer gap left by Fable 5’s troubled rollout.
The benchmark data is striking. On Frontier-Bench v0.1, an agentic coding test scoring real software tasks end‑to‑end, Opus 5 hit 43.3% versus Fable 5’s 33.7% and OpenAI’s GPT‑5.6 Sol at 34.4%. More dramatic is ARC-AGI‑3, a novel problem‑solving benchmark: Opus 5 scored 30.2% while GPT‑5.6 Sol reached just 7.8%. On the knowledge‑work Elo rating GDPval‑AA v2, Opus 5 scored 1,861 against Fable 5’s 1,747 and GPT‑5.6 Sol’s 1,736. In practical terms, Zapier’s AutomationBench saw Opus 5 complete a full churn‑prevention workflow with a 100% pass rate—something no prior model had achieved. Devin CEO Scott Wu noted the model “approaches Fable‑level performance at half the cost,” while Lovable’s Fabian Hedin praised it as “steadier, with far less variance run to run” on the hardest coding tasks.
Anthropic’s lineup now runs in four tiers: Haiku (fast/cheap), Sonnet (mid‑range), Opus (workhorse), and the restricted Mythos class that includes Fable 5 and the super‑restricted Mythos 5. Fable 5 had a rocky launch in June: pulled after an emergency export order due to a jailbreak vulnerability, then returning as a credits‑only model. Opus 5 now takes on the everyday heavy‑lifting role Fable 5 couldn’t hold, while offering 85% fewer safety classifier firings and a new Automatic Fallbacks feature that routes blocked requests to weaker models. The model shines in specialized domains too—Ultima Genomics reported it “behaves more like a careful scientist” in life‑science analysis—though it still lags behind Mythos 5 on cybersecurity tasks. The launch caps a week after Beijing‑based Moonshot AI released Kimi K3, an open‑weight 2.8‑trillion‑parameter model, but Opus 5’s price‑performance ratio firmly resets the enterprise AI landscape.
