xAI has launched Imagine Image 2.0 as a new Quality Mode on grok.com/imagine and within its iOS and Android applications. The model secures second place in global Arena benchmarks as of 7 August 2026, trailing OpenAI‘s GPT-Image-2 by specific Elo ratings. In the Image Edit Arena, Imagine Image 2.0 scores 1,439 against a 1,463 rating for the competitor. It also records 1,320 in the Text-to-Image Arena compared to 1,380 for GPT-Image-2. This performance places it ahead of other major contenders such as Meta’s Reve 2.1, Alibaba’s Qwen-Image-3.0-Pro, Google’s Gemini, and ByteDance’s SeedDream. The update significantly improves upon the previous Quality variant of the model.
The release focuses on instruction following and visual consistency for complex tasks. Users can access editing tools like a Magic Wand for selective modifications and a segmentation feature for precise region selection. A background removal tool exports subjects with transparent backgrounds. Multi-Ref Editing allows combining up to five input images into a single generation. Smart Resize converts existing images to any aspect ratio while filling extra space automatically. The system also offers preconfigured templates for photo editing, product photography, and game assets. These features aim to support iterative workflows and video production planning.
* API access for Imagine Image 2.0 is scheduled for release soon
* The model beats the predecessor’s Quality variant by a wide margin in tests
* Visual style remains consistent across generated characters, locations, and props
