Anthropic launches ClaudeVision 2 — a multimodal assistant built for creative teams

AI · 5 min read

Anthropic’s ClaudeVision 2 is a multimodal model and application layer aimed at designers and creative teams, combining image understanding, generative editing, and a new API that hooks directly into design tools. The company says the model runs inference in hybrid cloud configurations, allowing sensitive assets to stay on-prem while leveraging Anthropic’s hosted compute for heavier tasks.

ClaudeVision 2 introduces contextual layer editing: designers can point at a part of a mock and ask the agent to generate variants, match brand tokens, or produce component code for React and Framer. Anthropic emphasized latency improvements and a 40% reduction in prompt engineering overhead for iterative design tasks compared to previous tools.

Customers in the beta include three ad agencies and two large retailers. Anthropic also revealed enterprise pricing tiers with SLAs and a self-hosted runtime option aimed at agencies that handle high volumes of proprietary creative assets.