Claude Opus 5 Is Highly Capable, But Is No Mythos

Anthropic released Claude Opus 5, and AI researcher Zvi Mowshowitz offers a detailed technical take. Opus 5 is positioned as the more affordable option next to Fable 5—half the price per token, with most of Fable's everyday performance. Benchmarks are strong, especially in coding and math, and notably, it's harder to prompt-inject than previous Claude versions. But it has clear limits. Opus 5 is designed as a subagent, not the lead model. It shines at bounded, well-defined tasks but struggles with autonomous, multi-step reasoning. Per Mowshowitz, it lacks what he calls 'the Juice'—the ability to connect disparate ideas across domains. And several users report personality friction: they find it repetitive, prone to sprawling sentences, sometimes paranoid or argumentative. If you enjoyed Opus 4.7 or 4.8, you'll likely click with Opus 5. If not, Fable remains the better choice.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton