Grok Imagine Image 2.0 Ships: xAI's Typography-Aware Image Model Is Already on Vercel's AI Gateway
TL;DR
Grok Imagine Image 2.0 is xAI's typography-aware image model, ranked second on text-to-image and editing leaderboards and callable via Vercel's AI Gateway.
Last updated: August 8, 2026
xAI shipped Grok Imagine Image 2.0 on August 7, 2026, and it is the first image model from a frontier lab built around a developer-shaped problem: text. The model plans typography and layout before it paints, so dense multi-part visuals like infographics, posters, and title screens hold their structure and small text stays legible. xAI says 2.0 ranks second in the world in both text-to-image generation and image editing on the Arena leaderboards (Elo, as of August 7, listed under SpaceXAI).
The developer-facing detail: a 2.0 preview build is already on Vercel's AI Gateway as xai/grok-imagine-image-2.0-preview, callable from the AI SDK today - even though xAI's own announcement says API access is "coming soon." Gateway first, vendor API second: that ordering is the story for anyone who builds image features.
Official Sources#
| Resource | Description |
|---|---|
| xAI announcement: Imagine Image 2.0 | The official release post, August 7, 2026 |
| xAI Imagine API docs | Model capabilities, request shapes, resolution tiers |
| xAI API pricing | Verified live pricing, August 8, 2026 |
| Vercel changelog: Grok Imagine Image 2.0 on AI Gateway | Model id, AI SDK usage, playground |
| AI SDK generateImage docs | The call shape used below |
What shipped#
Imagine Image 2.0 is generally available as the new Quality Mode on grok.com/imagine and the iOS and Android Grok apps. xAI frames it around one goal: "make images you can use in real work." Concretely, that means three capability groups:
- Instruction fidelity with typography planning. 2.0 plans text layout the way a designer would, before generating pixels. The result is that infographics, posters, and other text-dense outputs keep their structure, and small type renders sharp instead of mushy. This is the output xAI leads the announcement with:
-
Precise editing as a first-class feature. The consumer surface adds a magic wand that edits only the region you point at, segmentation that selects exact areas, background removal that exports a subject with a transparent background, and multi-reference editing that accepts up to 5 input images in a single generation - no manual compositing step.
-
Smart resize. Give the model one image and a target ratio, and it recomposes the scene into the new frame instead of cropping it. Supported ratios cover 1:2, 9:16, 2:3, 3:4, 1:1, 4:3, 3:2, 16:9, and 2:1. The same persistence machinery keeps a character, location, or prop consistent across separate generations, which xAI positions as building "one world" for video pre-production:
The announcement also ships 15 templates (photo edit, product color change, e-commerce listings, headshots, icon maker, character sprite, merch maker, and more) that pre-configure a workflow so you supply inputs and get a finished asset.
The leaderboard claim#
xAI says Image 2.0 "ranks second in the world in both text-to-image generation and image editing," citing the Image Edit Arena and Text-to-Image Arena leaderboards as of August 7, with xAI listed under its parent company name, SpaceXAI. That is the vendor's own framing, not an independent measurement, and "second" hides who is first - the numbers on the leaderboards themselves are the source if you want the full table. For developers the meaningful claim is directional: a frontier text model vendor has decided text rendering and structured layouts are the competitive battleground for image generation.
Pricing, verified August 8, 2026#
xAI has not published 2.0-specific API pricing yet - the API access is still "coming soon" - so the current Imagine API lineup is the reference until it lands:
| Model | Input | Output 1K | Output 2K |
|---|---|---|---|
| grok-imagine-image-quality | $0.01 / img | $0.05 / img | $0.07 / img |
| grok-imagine-image | $0.002 / img | $0.02 / img | $0.02 / img |
Source: the xAI pricing page, fetched today. Image generation is flat per-image regardless of prompt length, and edits are billed for both the input image and the generated output. A single request can return up to 10 images. The quality tier sits at $0.05 per 1K image, which is the tier 2.0 will presumably replace when its API pricing lands.
How to call it today#
The 2.0 preview is on Vercel's AI Gateway, and the changelog shows the exact AI SDK shape:
import { generateImage } from 'ai'
const { images } = await generateImage({
model: 'xai/grok-imagine-image-2.0-preview',
prompt: 'An infographic tracing letterforms from movable type to digital fonts.',
})
Resolution is set per call - 1k or 2k under providerOptions.xai - and n controls how many images come back. Editing works by passing an image in prompt.images alongside the instruction, so the model changes what you asked for and leaves the rest. Vercel also hosts a playground at imagine.vercel.sh running on the gateway, which is the fastest way to feel the difference 2.0 makes on text-dense prompts without writing code.
One honest caveat: this is an image model, not a text model, so it does not run in OpenCode or any coding agent - the agent-side integration is the standard pattern of calling generateImage from tool code. If you want the text-side xAI experience, Grok 4.5 is the one to wire into your coding agent.
Why it matters#
Image models have been excellent at single-subject aesthetics and unreliable at anything with words in it, which is exactly what production assets - social cards, product shots, game UI, documentation diagrams - are made of. A model that plans layout before rendering and can edit precisely after, at roughly $0.05 per 1K image, moves image generation from "prompt for a hero image" toward "actually produce the asset." The gateway-first availability through Vercel is the same pattern we saw with FLUX 3's rollout: the fastest route to a new model increasingly runs through the platform layer, not the vendor's own SDK, and Vercel has been collecting those routes for image and video models alike - Meta Muse, MiniMax H3, and now Grok Imagine 2.0.
For production image workflows, pair the model with prompt discipline: OpenAI's GPT-Image 2 prompt library shows the difference structured prompting makes on text-bearing output, and that lesson transfers directly. If 2.0 really holds small type and layout structure at $0.05 per image, the next generation of infographic, poster, and title-screen generation is going to be built on prompts like the one above.
FAQ#
What is Grok Imagine Image 2.0?#
xAI's next image generation model, announced August 7, 2026 and GA as the new Quality Mode on grok.com/imagine and the Grok mobile apps. It emphasizes instruction fidelity, typography and layout planning, subject consistency, and precise editing tools.
When is Grok Imagine Image 2.0 available through the xAI API?#
The official announcement says API access is "coming soon." A 2.0 preview build is already available now through Vercel's AI Gateway as xai/grok-imagine-image-2.0-preview, callable with generateImage from the AI SDK.
How much does Grok Imagine Image 2.0 cost?#
2.0-specific API pricing is not published yet. The current quality-tier model, grok-imagine-image-quality, costs $0.01 per input image and $0.05 per 1K (or $0.07 per 2K) output image, with flat per-image pricing and edits billed for input plus output.
Can I run Grok Imagine Image 2.0 in OpenCode?#
No. It is an image generation model, not a text model, so it does not slot into coding agents. The standard integration is calling generateImage from tool code via the AI SDK or the xAI API once access goes live.
Continue Reading#
- FLUX 3: Black Forest Labs Ships a Unified Multimodal Foundation Model - the other August image-model release, spanning image, video, and audio
- Meta Muse Developer Guide - how Meta's image and video generation model compares
- GPT-Image 2 Prompt Library: What Works in Production - structured prompting for text-bearing image output
- MiniMax H3 Omni Video Model - the video side of the media-generation wave
- Grok 4.5 for Developers - the xAI text model worth wiring into your coding agent
- GPT Image 2.5 Sunburst and Flare - OpenAI's new image models, the xhigh and max quality tiers, and how to estimate cost
- Image and Video Generation API Pricing in October 2026 - Grok Imagine Image 2.0 against every other per-image price, and where xAI's markdown and rendered pricing pages disagree
Sources#
- xAI: Imagine Image 2.0 announcement, August 7, 2026
- xAI API pricing page, verified August 8, 2026
- xAI Imagine API docs
- Vercel changelog: Grok Imagine Image 2.0 on AI Gateway, August 8, 2026
- AI SDK generateImage reference
- Arena leaderboards - the leaderboard xAI cites for its second-place claims
Get the next deep dive like this in your inbox
One email a week on News and the rest of the AI dev stack. Free.
Read next
Grok 4.5: xAI Releases Cursor-Trained Coding Model at $2/M Input Tokens
Grok 4.5 coding model from xAI, trained on Cursor data: $2/M input pricing, benchmarks near Opus 4.7, the CursorBench caveat, and the Hacker News reaction.
5 min readGrok 4.6: xAI's Agent-Focused Update Matches GPT-5.6 Sol at the Same $2/$6 Price
Grok 4.6 release guide: xAI's agent-focused update matches GPT-5.6 Sol on the AA index at the same $2/$6 pricing, and runs in Cursor and OpenCode.
8 min readFLUX 3: Black Forest Labs Ships a Unified Multimodal Foundation Model for Image, Video, Audio, and Robotics
FLUX 3 from Black Forest Labs is one multimodal model for image, video, audio, and robot actions. What it does, how it works, and how to try it.
10 min readNew here? Start with
Technical content at the intersection of AI and development. Building with AI agents, Claude Code, and modern dev tools - then showing you exactly how it works.
