Claude Design Launch: Opus 4.7 Powers AI UI Generation

Claude Design converts Figma files into prototypes, pitch decks, and production‑ready user interfaces, eliminating the need for traditional design software. The launch sent Figma’s stock down 7%, signaling a disruptive shift in the design industry. Professionals who once focused on UI layout now face a transition toward prompt engineering roles.

Technical Specifications of Opus 4.7

Opus 4.7, the model powering Claude Design, boasts a higher “taste” and creativity rating than its predecessor. Vision capabilities handle images up to 3.75 megapixels (2576 pixels on the long edge), a notable upgrade for design work. On a software‑engineering benchmark the model scores 87.6%, beating Opus 4.6 but still trailing the mysterious Mythos model. Some users allege a regression in performance, fueling speculation about intentional “nerfing” of the earlier version.

Functional Capabilities

The platform generates fully interactive UIs complete with out‑of‑the‑box animations. It can also produce complex shader‑based animations and video content exceeding one minute in length. Design‑system consistency is maintained by linking a GitHub repository or uploading a Figma file, allowing the model to reference existing style guides.

Hands‑on Evaluation

In direct testing, Opus 4.7 proved slower than competitors such as Google Stitch, Codeex, or Cursor Composer, though it remains orders of magnitude faster than manual human effort. The tool sometimes struggles to adhere strictly to uploaded design systems, resulting in minor mismatches like subtle background‑color changes. An interactive canvas lets users draw arrows, add comments, and issue prompts for iterative fixes; each iteration typically consumes five to ten minutes and does not always achieve the desired outcome.

Mechanisms & Explanations

Design‑system integration works by ingesting a GitHub repo or Figma file and using it as a reference point for visual consistency across generated screens. The interactive canvas editing feature translates on‑screen annotations into model updates, enabling a feedback loop that refines the UI. Deployments can leverage Google Cloud Run’s serverless infrastructure, which scales to zero when idle and auto‑scales during traffic spikes, supporting multiple languages and frameworks.