Grok Imagine Image Pro

Grok Imagine Image Pro is xAI's premium image generation model, built on the same Aurora autoregressive mixture-of-experts architecture as the standard variant but tuned for higher fidelity output and superior editing performance. Aurora was trained on billions of examples of interleaved text and image data, giving it native multimodal understanding that enables both generation from scratch and intelligent editing of existing images. The Pro model is particularly strong in single-image editing tasks, where it ranked among the top models on Arena.ai's single-image-edit benchmark with a leading score.

The model supports text-to-image generation and image editing with output at up to 2K resolution — double the resolution of the standard Grok Imagine Image model. This makes it ideal for use cases requiring fine detail and crisp output, such as high-resolution concept art, detailed product renders, and production-quality assets. In editing mode, provide a source image with a natural language prompt describing your changes, and the model will intelligently modify the image while preserving its structure and style. You can select from multiple aspect ratios (2:1, 16:9, 3:2, 4:3, 1:1, 3:4, 2:3, 9:16, 1:2, or auto) and generate up to 10 images per request for rapid creative exploration.

Grok Imagine Image Pro sits on the Pareto frontier of Arena.ai's Image Arena for cost-to-quality in the 2-8 cents per image range, meaning it delivers the highest benchmark scores at its price point. The model excels across diverse visual styles — photorealism, illustration, painting, and stylized art — with precise prompt adherence and accurate rendering of text, fine details, and complex compositions. With support for prompts up to 10,000 characters, you have full creative control over scene descriptions, lighting, style, and composition direction.