xAI has introduced Imagine Image 2.0, a new image generation and editing model available as a Quality Mode on grok.com/imagine and inside Grok’s iOS and Android apps. The release puts xAI close to the front of current image AI rankings, with the model landing behind OpenAI’s GPT-Image-2 in Arena benchmarks.
The update is not only about generation quality. xAI is also packaging Imagine Image 2.0 with editing tools, templates, and workflow features aimed at people who need to revise, resize, combine, and prepare visuals rather than create a single image and stop there.
A strong benchmark debut
In the Arena leaderboards as of August 7, 2026, the faster "low" variant of Imagine Image 2.0 takes second place globally in both tested categories mentioned in the source article. That result places it directly behind OpenAI’s GPT-Image-2 in both image editing and text-to-image generation.
In the Image Edit Arena, Imagine Image 2.0 scores an Elo rating of 1,439. OpenAI’s GPT-Image-2 leads that category with 1,463. In the Text-to-Image Arena, Imagine Image 2.0 reaches 1,320, while GPT-Image-2 is ahead at 1,380.
The same leaderboard context puts several other image models further down the list, including Reve 2.1, Meta’s Muse-Image, Alibaba’s Qwen-Image-3.0-Pro, Google’s Gemini, and ByteDance’s SeedDream. The new xAI model also performs well above the "Quality" variant of its predecessor in these tests.
For users comparing image AI tools, the ranking matters because it suggests Imagine Image 2.0 is competitive across two different jobs: changing existing images and creating new ones from text. Those are related tasks, but they are not identical. A model can be useful for one and less effective at the other, so appearing near the top of both categories is a notable result within the facts provided.
What xAI says the model is built to do
xAI describes Imagine Image 2.0 as a model designed for accurate instruction following, clean typography and layout in complex visuals, and consistency across multiple generations. Those capabilities matter in practical image work because visual tasks often depend on details: where something appears, how text is arranged, and whether a series of outputs still looks connected.
Fine-grained instruction following can help when a user wants a specific edit rather than a broad reinterpretation of the whole image. Clean typography and layout are important when generated visuals include structured elements or readable text. Consistency across multiple generations becomes more relevant when users are creating a set of assets instead of one isolated picture.
The model is currently available through grok.com/imagine and Grok’s mobile apps. xAI says API access for Imagine Image 2.0 is coming soon, which would extend the model beyond the current product surfaces once that access arrives.
Editing tools for repeated revisions
Imagine Image 2.0 includes several editing features that point toward iterative workflows. Instead of treating image generation as a single prompt-and-output process, the toolset gives users ways to select, isolate, combine, and expand images after the first result appears.
One feature is called "Magic Wand." According to xAI, it changes only the selected area of an image. A segmentation feature lets users choose precise regions, while a background removal tool can export subjects with a transparent background.
Those tools cover common editing needs:
- Selecting one part of an image and changing only that region.
- Choosing precise visual areas before applying an edit.
- Removing a background so the subject can be reused elsewhere.
The release also includes Multi-Ref Editing, which lets users combine up to five input images into one generation. That feature is aimed at cases where a single reference is not enough. A user may need the subject from one image, the setting from another, and visual details from additional inputs, all brought together into a new result.
Another feature, "Smart Resize," converts an existing image to any aspect ratio. The model fills in the additional space on its own. That can be useful when one visual needs to fit multiple formats without simply cropping away important parts of the image.
Templates broaden the use cases
xAI is also adding templates that turn common image tasks into preconfigured starting points. The categories named in the source article include photo editing, product photography, marketing materials, design tools, game assets, and streaming emojis.
Templates can reduce setup time for users who return to the same kinds of work repeatedly. Instead of building every prompt or editing sequence from scratch, they can begin from a structure that already matches a common purpose.
The category range also shows that xAI is positioning Imagine Image 2.0 for several different creative and commercial contexts. Product photography and marketing materials point toward promotional visuals. Game assets and streaming emojis suggest smaller, repeatable graphics. Photo editing and design tools cover broader visual production needs.
A step toward video planning
Beyond still images, xAI is showing a feature that creates characters, locations, and props separately while keeping the visual style consistent across all images. The company frames this as a stepping stone toward full video production workflows.
The important point is consistency. When characters, environments, and objects are generated separately, they still need to belong to the same visual world. If their appearance changes too much from one output to the next, the assets become harder to use together.
Based on the source article, Imagine Image 2.0 is best understood as both a model upgrade and a workflow expansion. Its Arena benchmark position puts it just behind OpenAI’s GPT-Image-2, while its editing tools, templates, and pre-production features show xAI trying to make image AI more useful after the first generation is complete.