While everyone is watching only the LLM performance race, xAI's image model has quietly climbed to No. 2 in the world. What Grok's official account said is simple: Imagine Image 2.0 is No. 2 on Arena in both text-to-image generation and image editing.
Right Behind OpenAI in Both Categories, by the Numbers
| Category | No. 1 | No. 2 (xAI) | Gap |
|---|---|---|---|
| Text-to-Image Arena | gpt-image-2 (OpenAI) 1380 | grok-imagine-image-2 low 1320 | 60 points |
| Image Edit Arena | gpt-image-2 (OpenAI) 1463 | grok-imagine-image-2 low 1439 | 24 points |
In both categories, OpenAI's gpt-image-2 is No. 1, and xAI is right behind it. In image editing in particular, the gap is only 24 points, so it is effectively a close second.
A Detail to Check: "No. 2" Is for the Low Option
There is one point to be precise about. xAI submitted the same model to Arena with two options (low and quality), and the one that took No. 2 is the low option. The quality option scored 1228 in text-to-image, near the bottom of the leaderboard, and 1390 in image editing, around fifth. The "No. 2" in the announcement reads as if the whole model is No. 2, but the ranking actually splits widely depending on the option.
Three Precision Editing Tools Also Arrived
In the same announcement thread, xAI also released three new editing tools.
1. Magic Wand: change just one area of an image and keep the rest as is
2. Segmentation: precise area selection
3. Smart Resize: adjust to the aspect ratio you want
It was pitched for making infographics, ad images, game assets, UI/UX mockups, and storyboards, and an API is coming later.
No. 2 on Arena: Will It Win Over Real Work Too?
1. Having fallen behind in the LLM race, it made its presence felt in images. One reaction in the announcement thread said, "Grok 4.6 missed its release schedule, but Image 2.0 quietly became No. 2 in the world." The reading is that xAI got results in image models while the LLM core race is fierce.
2. It explicitly stressed practical uses. Using the phrase "real world usefulness" and naming work products like infographics, ads, and UI mockups reads as an intent to position it as a work tool, not just an image generation demo.
Arena Scores Are a Reference; Verify Yourself Until the API Opens
Arena rankings are widely used as a benchmark based on user votes, but they do not fully match real work quality. With the API not yet open, the surest approach at this point is to test it directly at grok.com/imagine.
Related: Grok's official X post