According to xAI’s announcement, Grok’s image models now trail just one rival on both major public leaderboards: OpenAI’s gpt-image-2. The standings, dated August 7, accompany Grok Imagine Image 2.0, released the same day as Quality Mode on grok.com/imagine and in the company’s mobile apps.
The update is built around editing tools that work on selected regions rather than whole images: a magic wand that alters only the area a user points at, segmentation for precise selection, and background removal that exports subjects with transparency. Multi-reference editing folds up to five input images into a single generation, and a smart-resize feature re-composes one image across nine aspect ratios, from tall 1:2 banners to wide 2:1 formats. Templates bundle common jobs such as product shots, headshots, e-commerce listings, game assets, and marketing posters, so users supply only the inputs.
Training emphasized visual fidelity across photos, illustrations, and design work, with typography planned as part of the layout so complex scenes stay legible, per xAI. The Arena entries appear under the SpaceXAI name, the corporate identity the parent company has adopted since its June Nasdaq listing. The release is consumer-first for now, with xAI listing API access as coming soon and giving no timeline.
