Anthropic's Claude Visual v2 debuts with UI-aware image generation for designers

AI · 5 min read

Anthropic today launched Claude Visual v2, a multimodal model iteration focused on UI-aware image generation and layout-sensitive outputs. The model includes a new internal representation that preserves component boundaries and spacing, which Anthropic says reduces the need for post-generation cleanup when creating interface assets.

Alongside the model release, Anthropic introduced an official Figma plugin that translates design tokens and constraints into structured prompts, allowing designers to generate button states, icons, and hero images directly within their files. The plugin supports responsive breakpoints and exports assets with consistent spacing and layer naming conventions.

Early adopters report faster iteration when exploring visual directions, but note occasional mismatches in accessibility attributes and color contrast. Anthropic has opened a feedback channel for designers to submit failing cases and plans weekly updates focused on accessibility and token fidelity.