Anthropic's Opus 5 slashes price by 50% and drops a controversial data policy

Anthropic just made a move that’s going to ruffle some feathers at OpenAI and Google. The company debuted Opus 5, its latest large language model, and priced it at exactly half of what its Fable sibling costs. That’s an aggressive cut in a market where inference costs are the real battlefield, not just benchmark scores.

The pricing isn’t the only story here. Opus 5 ships without a data retention requirement — meaning enterprises that squirm at the thought of their prompts living on someone else’s server can breathe a little easier. For regulated industries, this detail alone might matter more than any marginal improvement on a math benchmark. Anthropic is clearly betting that privacy-sensitive buyers have been sitting on the sidelines, waiting for a model that doesn’t force them to choose between capability and compliance.

What’s less clear is how Opus 5 actually performs. Anthropic’s announcement focuses on the economics and the policy shift, not a barrage of charts showing it crushing the competition. That silence is interesting. If the model were a knockout on quality, you’d expect them to lead with it. Instead, the pitch is about value and trust. In a market where every vendor claims their new model is the smartest, Anthropic is trying to compete on a different axis entirely.

The half-price move also puts pressure on the broader ecosystem. If a frontier lab can profitably serve a model at this price point, what does that say about the margins everyone else is enjoying? Expect enterprise procurement teams to start asking harder questions about why they’re paying what they’re paying.