Google DeepMind debuts Gemini Nano 2 for edge design assistants in Figma plugins

Tech · 4 min read

DeepMind quietly launched Gemini Nano 2, a smaller, multimodal model tailored for edge and mobile use cases. Designed for low-latency inference and modest memory footprints, the model allows plugin developers to run inference locally on laptops and higher-end mobile devices.

Early adopters in the Figma ecosystem have already shipped plugins that use Gemini Nano 2 for tasks like instant symbol extraction, offline image-to-vector tracing, and personal-brand-consistent palette suggestions. The local-first approach appeals to designers concerned about sending proprietary product screenshots to cloud services.

DeepMind framed the release around privacy and responsiveness, offering optional federated learning hooks for teams that want continuous improvement without centralized data capture. Pricing and licensing are aimed at platform partners and enterprise devs rather than individual consumers.