Opus 5 costs a third of the price — and that’s actually the problem - The New Stack

What the report says

The New Stack reports that Anthropic’s Opus 5 is being positioned as a high-performing model for coding work at a sharply lower price point, with the headline framing it as costing about one-third as much as a competing or referenced “Fable 5.” The supplied snippet says Opus 5 offers state-of-the-art coding performance at a fraction of that cost, making price a central part of the product’s significance rather than just a procurement detail.

The article’s main concern, as described in the available evidence, is that cheaper, more capable coding models could accelerate the use of autonomous software-development agents. Those systems can plan tasks, write or modify code, run commands and iterate with less direct human involvement than earlier AI coding assistants. The New Stack’s framing suggests the cost drop is “the problem” because it may encourage broader and longer-running agent workflows before organizations have mature controls in place.

The reported issues include security, spending and oversight. For engineering teams, lower per-use costs can still turn into material expenses if agents run many tests, call tools repeatedly or operate across large codebases. Security risks may also rise when agents are allowed to execute code, access repositories, handle credentials or interact with production-like environments.

More broadly, the development reflects a shift in AI coding from chat-style assistance toward agentic systems that need governance similar to other automated infrastructure. Based on the supplied material, the takeaway is not simply that Opus 5 is cheaper, but that lower-cost autonomy may require stricter budgets, permissions, audit trails and human review.