Histoire
juillet 31, 2026
Anthropic’s Opus 5 Makes Fable 5 Look Expensive
Anthropic has launched Claude Opus 5, its newest large-scale AI model, which it claims offers performance comparable to its most powerful model, Fable 5, but at half the price. The release emphasizes cost-efficiency and performance on tasks like coding, positioning it as an everyday tool for enterprise users.
Anthropic has a new AI problem of its own making: when your cheaper model looks nearly as good as the prestige one, the premium tier starts to feel a little awkward. That is the tension hanging over the launch of Claude Opus 5.
Anthropic’s pitch is straightforward. Opus 5 is supposed to deliver performance close to Fable 5 at roughly half the price, while becoming the company’s everyday model for developers, enterprise users, and knowledge workers.1 On paper, that is a strong commercial story. In practice, it also raises an obvious question: if Opus 5 is close enough, why pay up for Fable at all?
That question runs through much of the outside coverage. Ars Technica argued this is fundamentally a cost play, not a genuine jump in capability, calling Opus 5 “about token efficiency, not a capability leap.”2 Axios framed it similarly, saying Anthropic built the model to offer performance near Fable “on many tasks at half the price.”3 The Verge also emphasized that Opus 5 “comes close to the capabilities of Claude Fable 5 in many domains,” especially around complex coding work, while landing in a politically sensitive moment after cybersecurity scrutiny from Washington.4
That scrutiny matters. Anthropic and several reporters stress that Opus 5 is safer and less prone to misuse than its bigger sibling, even if that means it trails Fable and Mythos on offensive cyber behavior.14 TechCrunch went further, noting Opus 5 is both “cheaper and less restrictive than Fable,” and likely the better fit for most use cases.5
Still, not everyone sees a smooth landing. Every’s early hands-on verdict was blunt: Opus 5 was “brilliant in flashes, frustrating in practice,” with the team complaining that it fought instructions and stumbled inside existing workflows.6 Yet market appetite looks real. Perplexity touted that Opus 5 “outperformed all but Fable 5, while being 57% cheaper.”
7
Put it together and Opus 5 looks less like a moonshot than a ruthless piece of product segmentation: good enough, cheaper, safer—and uncomfortably close to the flagship.