Anthropic launches Claude Opus 5, its most advanced AI model

Anthropic announced Claude Opus 5, the most advanced version of its model family to date. According to the company, the new model arrives with performance close to a state-of-the-art internal system called Fable 5 — a leap that appears mainly on three fronts: logical reasoning, code writing, and understanding nuance in long texts. These are exactly the tasks where previous generations stumbled.

The launch happens in the midst of the dispute between Anthropic, OpenAI, and Google over who delivers the most capable model. The priority of the moment in this market has ceased to be conversing well and has become making fewer errors: fewer logic failures in multi-step tasks and a larger context window, which is what allows working with entire documents at once. This is the demand of those using AI in production, within companies, and not just to answer questions.

For those learning AI in Brazil, a new model matters for a simple reason: a more capable tool changes what you can do with it. Tasks that previously went wrong — summarizing long documents in Portuguese, reviewing legal text, helping to program — tend to start working. And there's an indirect effect that usually matters more: a launch of this caliber pushes previous models to cheaper or free plans, which expands access for those who can't afford a subscription in dollars.

Two examples of what changes in practice. An 80-page public notice from which you need to extract deadlines and requirements: the reading tends to stay more faithful to the original, with less invention and better organization. And hundreds of customer reviews scattered across different spreadsheets: you can load everything and ask not just for the summary, but for the pattern behind it — connecting a logistics complaint with the sales opportunity it hides.

The practical advice is the same as with every launch: don't believe the announcement, test it. If you already use Claude, repeat the tasks you do today on the new model and compare, especially in Portuguese. If you prefer another chatbot, create a free account and run the same test on both — the difference between models only becomes clear in real use, with your own material. Pay special attention to multi-step instructions, which is where the generation gap shows up first.

Two caveats about money. If you're a developer planning to use the API, the Opus line models are the most expensive in the family — it's worth measuring the gain before switching the model for an entire project. And if you're a user, wait a few months before subscribing to anything: a launch of this scale usually triggers adjustments from competitors, and what's paid today tends to drop in price.