Anthropic released Claude Opus 4.7 with claimed improvements in complex software development, instruction following, and verification of generated work. Its visual capabilities support higher-resolution inputs, with examples involving interfaces, documents, and presentations.

Context

Noticing a mistaken assumption matters as much as producing more output. Model-led verification can add a useful check, but consequential changes still benefit from independent tests and a clear account of what was examined. A claim that something was checked should be backed by observable results.

Sources & authors

  1. Introducing Claude Opus 4.7
    Anthropic · April 16, 2026