--- title: "xAI's grok-imagine-image-quality Hits Top 5 at 75% Less Than GPT Image 2" subtitle: "xAI's grok-imagine-image-quality lands at #5 on both major image leaderboards, undercutting OpenAI and Google on price while closing the quality gap" url: "https://alphasignal.ai/news/xai-s-grok-imagine-image-quality-hits-top-5-at-75-less-than-gpt-image-2" author: "AlphaSignal Newsroom" author_url: "https://alphasignal.ai/editorial-team" company: "Artificial Analysis" company_url: "https://alphasignal.ai/company/artificialanlys" published: "2026-05-28T18:59:36.000Z" updated: "2026-06-08T19:44:00.573Z" type: "News" topics: - "Image" - "Benchmarks" subtopics: - "Image Generation" - "Image Editing" upvotes: 162 read_time: "5 min" source_url: "https://artificialanalysis.ai/image/leaderboard/text-to-image" comments: 0 --- # xAI's grok-imagine-image-quality Hits Top 5 at 75% Less Than GPT Image 2 > xAI's grok-imagine-image-quality lands at #5 on both major image leaderboards, undercutting OpenAI and Google on price while closing the quality gap The AI image generation race just got a new serious contender. xAI has launched [grok-imagine-image-quality](https://x.ai/news/grok-imagine-quality-mode), a higher-fidelity variant of its Grok Imagine model, and it has debuted at #5 on both the Artificial Analysis Text-to-Image and Image Editing leaderboards. That makes it the strongest image model outside of OpenAI and Google -- and at a fraction of the price. ## Where it sits in the pecking order The Artificial Analysis leaderboard ranks models using an Elo rating system derived from millions of blind head-to-head votes in their Image Arena -- users compare two images from the same prompt without knowing which model made which. The current top five by Elo are: GPT Image 2 (high) at 1339, GPT Image 1.5 (high) at 1266, Nano Banana 2 (Gemini 3.1 Flash Image Preview) at 1260, Cosmos3-Super-Text2Image (agentic) at 1240, and Nano Banana Pro (Gemini 3 Pro Image) at 1219. grok-imagine-image-quality sits just outside that published top five, making it the leading model from any lab other than OpenAI or Google. To understand why this matters, you need context on what happened before it. GPT Image 2 launched on April 22, 2026, and immediately hit an Arena Elo of 1,512 -- a 241-point gap over every competitor. The Text-to-Image leaderboard had been one of the most competitive charts in AI for the past year. Google's Nano Banana Pro held the top spot for months, with Flux 2, Grok Imagine, and Reve V1.5 all within striking distance -- the top fifteen models separated by roughly 130 points total. GPT Image 2 didn't just win; it reset the benchmark entirely. grok-imagine-image-quality is xAI's answer to that new bar. ## What actually changed xAI released the Grok Imagine Quality Mode API, a quality mode for image generation and editing available to enterprise developers and teams, with a focus on higher realism, stronger text rendering, and better creative control. Under the hood, Grok Imagine uses xAI's proprietary Aurora model, which was designed specifically for visual content generation -- distinct from the earlier Grok 2 image feature, which relied on Black Forest Labs' Flux model. The three headline improvements xAI is emphasizing: - **Realism:** Realism is improved, with official examples emphasizing natural skin, material details, lighting, scene atmosphere, and photographic texture -- things that matter for commercial images. - **Text rendering:** The model now handles in-image text with significantly more accuracy, a historically weak area for most image generators. - **Creative control:** Using grok-imagine-image-quality, you can request six documented style transfers: oil painting with impressionist technique, pencil sketch, pop art, anime, watercolor, and ultra-realistic photography. The model supports 14 distinct aspect ratios and resolution options of 1K or 2K. The Imagine API lets you generate and edit images and videos with Grok Imagine models, supporting image generation, image editing with up to 3 reference images, video generation from text or still images, and video editing. ## The price is the story Quality-tier image generation has been expensive. grok-imagine-image-quality changes that calculus significantly. Here's how it compares at 1K resolution: | Model | Price per 1K images | Max resolution | | --- | --- | --- | | GPT Image 2 (high) | $211 | 4K | | Nano Banana Pro (Gemini 3 Pro Image) | $134 | -- | | grok-imagine-image-quality (2K) | $70 | 2K | | grok-imagine-image-quality (1K) | $50 | 1K | | grok-imagine-image (standard) | $20 | 1K | The previous premium model, grok-imagine-image-pro, was deprecated as of May 15, 2026. All new requests should use grok-imagine-image-quality at $0.055/image, while the standard grok-imagine-image model remains available at $0.02/image for cost-sensitive workloads. That 4x price gap between the quality and standard tiers gives developers a meaningful routing decision: draft and iterate on the cheap model, ship final assets on the quality one. ## Enterprise is the real target The point of this update is not to create another generic text-to-image entry point. It is to put Grok Imagine into enterprise content production workflows: product images, marketing assets, ad variations, UGC-style content, brand visuals, and video generation. The multi-image editing capability -- accepting up to 3 reference images per request -- is particularly relevant here. You can feed it a product photo, a brand style guide image, and a scene reference, and get a coherent output that respects all three. xAI says pairing its latest image model with its video capabilities can support social media video assets, product showcasing, ads, and more. This is part of a broader multimodal push: xAI's Grok Imagine Video 1.5 launched shortly after and immediately claimed the top position on the Image-to-Video Arena leaderboard with a 52 Elo point jump over version 1.0, surpassing Seedance 2.0, HappyHorse 1.0, and Google Veo. The image and video models are designed to feed into each other. Getting started via the [xAI Imagine API](https://docs.x.ai/developers/model-capabilities/imagine) is straightforward and compatible with the OpenAI SDK: ``` from openai import OpenAI client = OpenAI( base_url="https://api.x.ai/v1", api_key="YOUR_XAI_API_KEY", ) response = client.images.generate( model="grok-imagine-image-quality", prompt="A collage of London landmarks in a stenciled street-art style", ) print(response.data[0].url) ``` ## Where it still falls short Independent benchmarks are more nuanced than leaderboard positions suggest. GPT Image 2 remains the gold standard for physical logic, complex prompt adherence, and professional workflows. Grok Imagine Image Quality offers impressive 2K resolution and vibrant aesthetics at a competitive price, but still struggles with advanced physics like fluid dynamics. GPT Image 2 demonstrated stronger instruction-following on spatial components in head-to-head tests. The competitive picture breaks down roughly like this: - **Choose grok-imagine-image-quality** if you need high-volume, high-resolution commercial assets and the price delta over GPT Image 2 matters at scale - **Choose GPT Image 2** for work where physical accuracy, complex multi-object spatial reasoning, or batch consistency across image sets is non-negotiable - **Choose grok-imagine-image (standard)** for drafts, prototypes, or any pipeline where $20/1K images is the right cost profile ## What it means for the field The image generation market is stratifying fast. OpenAI holds a commanding quality lead. Google is competitive on both quality and speed. But the core direction of Grok Imagine Quality Mode is to push image generation from "fun" toward "usable in enterprise production, ## Sources - [X post by @artificialanlys](https://x.com/artificialanlys/status/2060073523528581249) - [Grok Imagine Quality Mode API | xAI](https://x.ai/news/grok-imagine-quality-mode) - [Text to Image Leaderboard - Top AI Image Models | Artificial Analysis](https://artificialanalysis.ai/image/leaderboard/text-to-image) - [Grok Imagine Quality Mode API: xAI wants image generation inside enterprise workflows](https://www.knightli.com/en/2026/05/07/grok-imagine-quality-mode-api/) - [Does Grok xAI Actually Have Image Editing in 2026?](https://www.atlascloud.ai/blog/case-studies/grok-xai-image-editing-capabilities) - [GPT Image 2 Review: OpenAI's Best Image Model Ever Just Topped the Global Leaderboard](https://renovateqr.com/blog/gpt-image-2-chatgpt-images-2-review-2026)