Anthropic Just Made Its Best Model the Cheapest — And That's the Real Story

You can build a model that's almost as good as your flagship — and sell it for half the price — but doing it on purpose, with fewer restrictions to boot, isn't a pricing decision. It's a strategy.

Anthropic released Claude Opus 5 on July 24, and from one angle it looks like a straightforward upgrade: better at coding, stronger safeguards, the usual spec bump cadence. But the numbers don't tell the story the way the product positioning does. Opus 5 costs $5 per million input tokens and $25 per million output tokens — identical to Opus 4.8, half of Fable 5, and slightly less than OpenAI's GPT-5.6. It's less restrictive than Fable. It's being marketed as an "everyday model" for enterprise knowledge work and biology. And it's now available inside GitHub Copilot.

What Anthropic is really doing is splitting its product line into two tiers, and that separation matters more than any benchmark score.

Two models, one company, two strategies

Fable 5 — the flagship Mythos-class model — has been through a gauntlet this year. It drew the US government's scrutiny over cybersecurity concerns, was taken offline along with Mythos 5 for several weeks, then brought back with significantly stronger guardrails. The experience left a mark on both the model and the company's public posture.

Opus 5 is the response. It inherits the strongest safety architecture Anthropic has ever shipped — the company calls it "the most aligned Opus model and the least susceptible to being tricked into misuse" — but it carries none of Fable 5's political baggage. It's good enough for most tasks, cheap enough for everyday use, and free enough of restrictions that developers won't bounce off it when building real applications.

Anthropic spokesperson Danielle Ghiglieri confirmed to The Verge that the company continued coordinating with the Trump administration throughout Opus 5's development. "This includes Opus 5," she said — a line that reads as both transparency and a strategic hedge.

Who wins and who loses in the split

For enterprise customers, the calculus is immediately favorable. Opus 5 handles knowledge work and complex coding tasks at a price point that makes it practical for mass deployment inside an organization. The Verge reports it "comes close to the capabilities of Claude Fable 5 in many domains" and is "much better at complex coding tasks." In some benchmarks it actually outperforms the flagship.

For Anthropic, the two-tier structure solves a design tension that has haunted every model release this year. OpenAI's GPT-5.6 went through a narrow two-week trial limited to government-approved entities before wider release. Anthropic now has a model it can point regulators at (Fable 5, with the full safety apparatus) and a model it can point customers at (Opus 5, with the full developer experience).

Developers get a new practical option: a Fast Mode that doubles throughput at twice the price, and an automatic fallback feature that silently drops to a lower-tier model when Opus 5's safeguards decline a request — a pragmatic admission that refusal-is-safety isn't always what users need.

The market reads the message

The launch landed in a specific window: weeks after Anthropic's regulatory standoff with the US government, days after OpenAI's security incident dominated headlines. The timing suggests a company that has learned to stage its product news around the political calendar, not despite it.

TechCrunch notes that Opus 5 "will be both cheaper and less restrictive than Fable, likely making it preferable in most use cases." That's a remarkable statement about a model that costs half as much: the cheaper option isn't the compromise product. In many ways, it's the better product for the bulk of what enterprises actually do.

ZDNet's coverage framed it around developer economics: "near-Fable performance at Opus pricing" with "prompt-cache-friendly tool changes." GitHub Copilot integration sealed the practical message — this is a model meant to be used, not just benchmarked.

What the two-tier model means for the industry

Anthropic is betting that the market will pay for choice, not just capability. One model for the safety-conscious regulator; another for the speed-conscious developer. One that leads every news cycle with a political story attached; another that just runs well and costs less.

It's an approach that mirrors what enterprise software companies have done for decades — license tiers, feature gates, segmented pricing — but applied to the AI model itself. The model's intelligence isn't the product. The model's position in the market is.

The risk is fragmentation: the more models Anthropic maintains, the thinner its safety resources spread, and the harder it becomes to guarantee consistent behavior across the lineup. But for now, the bet is working. Opus 5 arrives not just as a capable model, but as the clearest signal yet that the AI market has entered its platform era — where how you sell matters as much as what you build.

Sources

• Anthropic launches Opus 5 — TechCrunch
• Anthropic releases Opus 5 with 'close' to Fable 5's capabilities — The Verge
• Claude Opus 5 arrives with near Fable performance at half the price — ZDNet
• Claude Opus 5 is now available in GitHub Copilot — The GitHub Blog
• Anthropic releases Claude Opus 5 for both AI coding and general office work — Fast Company
• Anthropic Releases Claude Opus 5 to Be Your New 'Everyday' Assistant — CNET