SpaceXAI 發表 Grok Imagine Image 2.0：精準編輯、Arena 雙料第二，直接挑戰 OpenAI

SpaceXAI has officially launched Grok Imagine Image 2.0, a significant upgrade to its image generation model, which is now available on grok.com/imagine as well as on iOS and Android applications. This upgrade follows the independent release of Grok Imagine 1.0 in February 2026 and represents a substantial evolution in the capabilities of xAI in the realm of image generation. The core focus of this update is clear: to elevate editing capabilities from being an 'add-on' feature to a 'primary function.' According to xAI's official statement on the X platform, Image 2.0 aims to 'help you create images that can truly be used in work,' highlighting four key pillars of this iteration: precise editing, clear text rendering, enhanced factual accuracy, and usability in practical application scenarios.

Enhanced Editing Capabilities

The most notable change in Grok Imagine Image 2.0 is the introduction of several precise editing tools. The Magic Wand tool allows users to select specific areas of an image for modification while leaving the rest untouched. The segmentation selection feature enables precise circling of areas that require adjustment, making it suitable for design scenarios that demand local fine-tuning. Additionally, the background removal function can directly output the subject with a transparent background, facilitating integration into other design software for further use.

A practical new feature is the multi-image reference editing capability, which allows users to input up to five reference images simultaneously. This eliminates the previous need for manual material stitching. For instance, designers can input product photos, brand color swatches, and scene images all at once to generate a product image that aligns with brand aesthetics. The smart resizing feature can recompose the same image into nine different aspect ratios, from a 1:2 vertical banner to a 2:1 wide banner, automatically filling in the new frame's content without the need for manual cropping or regeneration.

SpaceXAI has also introduced a template system that presets various common workflows, including product photography, professional headshots, e-commerce product images, game assets, and marketing posters. Users simply need to provide materials to quickly produce finished products, significantly lowering the entry barrier for those without a design background. Grok Imagine Image 2.0 can generate infographics, advertisements, game assets, UI/UX mockups, storyboards, and more, thereby broadening its applicability.

Competitive Positioning in the Market

According to data released by SpaceXAI, Image 2.0 has achieved impressive scores on the Arena rankings, securing 1320 points in the text-to-image category and 1439 points in the image editing category, both earning it the global second place. The only competitor ahead is OpenAI's gpt-image-2. In the Arena rankings, xAI's model surpasses major competitors such as Meta, Microsoft, Google, Alibaba, and ByteDance. The model is listed under the SpaceXAI name in Arena, reflecting a branding adjustment at the company level.

Technical Architecture and User Interaction

The underlying technology of Grok Imagine employs a dual-engine architecture. The Flux-based image stacking engine is responsible for generating and editing static images, known for its precise prompt adherence and clean text rendering. The Aurora engine, developed in-house by SpaceXAI, is a self-regressive model specifically designed for dynamic content, including image-to-video conversion and video editing, with native audio generation support. Image 2.0 emphasizes fidelity across photography, design, and illustration fields during training, optimizing layout and typography to ensure that complex visual designs with multiple elements maintain overall consistency and that small text is rendered clearly.

A distinctive design feature of Grok Imagine is its conversational iteration capability. Users do not need to write all their requirements in a single prompt; instead, they can adjust iteratively through multiple rounds of dialogue. For example, a user can first generate a base image and then instruct the model to 'change the sky to evening colors,' 'add a character,' or 'adjust the lighting to a warmer tone.' This conversational memory allows images within the same chat thread to maintain a consistent visual style, which is particularly useful for brand designers needing to produce a series of materials.

Subscription Plans and Accessibility

As of March 19, 2026, the image generation features of Grok Imagine have transitioned from free to paid subscription access. The subscription plans are as follows: X Premium at approximately NT$260 (USD $8) per month for basic image generation features; SuperGrok Lite at approximately NT$325 (USD $10) for basic advanced features; SuperGrok at approximately NT$975 (USD $30) for unlocking more generation quotas and full features; SuperGrok Plus at approximately NT$3,250 (USD $100) for a high-tier plan; and SuperGrok Heavy at approximately NT$9,750 (USD $300) for full feature access, including the latest models like Grok 4.5. Free accounts currently do not have access to image generation and editing features, with the system displaying a message indicating that these functions are limited to paid subscribers. The quota operates on a rolling reset mechanism rather than a daily midnight refresh, allowing quotas to gradually restore as older generation records expire.

In summary, Grok Imagine Image 2.0 represents a significant advancement in image generation technology, providing enhanced editing tools and a user-friendly interface that caters to both professional designers and casual users alike. As it positions itself against established competitors like OpenAI, its impact on the market will be closely watched by creators and technologists alike.