Anthropic releases Claude Opus 4.7, a new flagship for complex coding

Anthropic has released Claude Opus 4.7, the most powerful of its publicly available models to date. The company positions it as a notable step up from Opus 4.6, primarily for advanced software development and work with complex code.

The main change is in coding. Until now, Opus 4.6 handled routine tasks reasonably well, but in truly complex scenarios developers often had to guide the model closely, literally holding its hand and clarifying every step. According to Anthropic, Opus 4.7 needs such intervention far less often: it works out the context of a task on its own, determines more precisely what needs to be implemented, and gets past technical difficulties that tripped up its predecessor.

Beyond code, the update touches image handling. Opus 4.7 analyzes visual content more accurately: it distinguishes fine details, interprets charts and diagrams more correctly, and extracts text from screenshots and documents more reliably. The model has also improved at following multi-step instructions without losing detail at each step, and it holds the context of long chains of requirements better.

The third area of improvement is creative output. Anthropic says that when building presentations and documents, Opus 4.7 takes a more varied and unconventional approach to structuring material and choosing wording. For enterprise users who need text that is not merely formally correct but persuasive and well written, this may have real practical value, especially for automated generation of reports and business documentation.

The release comes amid active discussion of another Anthropic launch. A few weeks ago the company announced Mythos Preview, a model focused on cybersecurity that Anthropic itself calls the most powerful it has ever built. Mythos is not a standard commercial product: it is tailored to specific information security tasks and is so far available only in preview to a limited group of users. Opus 4.7, by contrast, is released openly as the main tool for a broad audience of developers, analysts and companies that use Claude through the API or a subscription.

The coexistence of two powerful models reflects an increasingly common industry trend: major AI companies are starting to ship separate lines, with general-purpose flagships for mass use and highly specialized systems for niche scenarios with extreme requirements. Anthropic, OpenAI and Google are moving in this direction in step. Such a split lets them avoid sacrificing specialized performance for the sake of versatility, since each line can be optimized independently.

The wording matters here: the most powerful generally available model is not the same as the most powerful model. Mythos Preview sits higher in Anthropic's internal hierarchy but remains a closed tool with restricted access. Opus 4.7 is the best the company is ready to offer a broad audience right now. For the vast majority of real-world tasks in development, analytics and document work, that level is more than enough.