Anthropic unveils Opus 5 with stronger coding, better efficiency, and lower cost. Here’s what changed and why the new model matters.

Anthropic Unveils Opus 5: What’s New and Why It Matters

Anthropic has released Claude Opus 5, its newest heavyweight AI model, and the headline is simple: it aims to deliver near-flagship performance at a much friendlier price point. The company says Opus 5 is cheaper and less restrictive than its top-tier Fable model, while still outperforming it on several benchmarks and handling coding and knowledge-work tasks with noticeably less back-and-forth.

What Changed

Opus 5 is positioned as a more practical everyday model for people who want strong capability without paying premium rates. Anthropic says pricing stays at $5 per million input tokens and $25 per million output tokens, unchanged from the previous Opus tier, even as performance improves. It also adds an “effort” control so users can choose low, medium, or high compute usage depending on how complex the task is.

Another notable shift is usability. Anthropic says Opus 5 is better at checking its own work, recovering from mistakes, and completing longer tasks with fewer corrections from the user. That makes it especially appealing for coding, research, and workflow automation, where constant prompting can slow everything down.

Benchmarks and Strengths

According to Anthropic’s announcement, Opus 5 sets a new internal high mark on coding and knowledge-work evaluations, including Frontier-Bench and GDPval-AA. The company also says it gets close to Fable 5 on many tasks, while costing less and being easier to access for regular users.

The model is also being pitched as Anthropic’s most capable generally available option for scientific research, with stronger biology-related performance than its predecessor. That matters because research users often need a mix of reasoning, precision, and long-context consistency rather than just raw conversational ability.

Safety and Access

Anthropic says Opus 5 is its most aligned Opus model yet and one of the hardest in its lineup to manipulate into misuse. At the same time, it is less restrictive than Fable and does not carry the same 30-day data retention policy that applies to some other models. For privacy-conscious users, that detail will likely matter just as much as benchmark gains.

Opus 5 is now the default model on Claude Max and the strongest model available to Claude Pro subscribers, according to reports. Anthropic appears to be aiming for a sweet spot: powerful enough for serious work, but not so locked down or expensive that it becomes niche.

Why It Matters

This launch is part of a wider trend in AI: the market is shifting from “biggest model wins” to “best model for the price.” Anthropic’s move suggests it wants Opus 5 to be the model people actually use day to day, while keeping Fable for the most demanding autonomous projects. In other words, Opus 5 is less about spectacle and more about being the sensible default.

For developers, researchers, and power users, that makes the release more important than it first appears. If Anthropic can really deliver near-frontier quality at a lower cost, Opus 5 could become the model many teams choose for coding, office work, and analytical tasks.