clauding.de
DE
• 2 min read AI-generated

Claude Opus 5 Is Here: Near Fable 5 Intelligence at Half the Price

Copy article as Markdown

Anthropic launched Claude Opus 5 — close to Fable 5's frontier intelligence, the company says, but at half the cost. Opus 5 becomes the new default on Max and Pro. And in one area, the model deliberately holds back.

Anthropic Claude Opus 5
Featured image for "Claude Opus 5 Is Here: Near Fable 5 Intelligence at Half the Price"

Anthropic shipped Claude Opus 5 on Thursday — and this time it’s less about a new benchmark record than about economics. The headline: nearly all the intelligence of Fable 5, at half the price.

What’s new

Opus 5 is available today in the Claude app and on the developer platform. Anthropic calls it a ‘thoughtful and proactive’ model and positions it as the go-to choice for most tasks — knowledge work, coding, automation, long-running agents.

The numbers behind it: Opus 5 beats Fable 5 on eight of thirteen benchmarks despite costing around 50 percent less. On coding and knowledge-work tests like Frontier-Bench and GDPval-AA, Anthropic claims new highs.

Pricing stays put: $5 per million input tokens, $25 per million output tokens — identical to its predecessor, Opus 4.8. What changed is the placement: Opus 5 becomes the default on Claude Max and the strongest model on Claude Pro. It shows up in Claude Code from version 2.1.219, and GitHub Copilot enabled it the same day.

The deliberate step back

One detail stands out: Anthropic stresses that Opus 5 is explicitly not state of the art for ‘risky, dual-use capabilities’ like cybersecurity. That’s not an oversight — it’s by design. If you’ve followed the last few weeks — open-weight models, sandbox escapes, misuse scenarios — you understand why a frontier lab mid-IPO would rather withhold a capability than advertise it.

A second point fits the same picture, and enterprises will like it: unlike Fable 5, Opus 5 doesn’t retain user data for 30 days. For companies with strict privacy requirements, that’s a concrete argument.

My take

The real trend isn’t the benchmark number, it’s the direction. A year ago ‘best model’ meant pricier, bigger, more. Opus 5 flips that — cheaper at almost the same performance. That’s exactly the lever that moves AI from demo to daily use. And the fact that Anthropic is intentionally braking on cybersecurity while heading into a record IPO is a telling signal: capability is no longer automatically the selling point. Sometimes restraint is.

Sources: