🇨🇳China·AI News·11 Nov 2026·via Tech Node

    MiniMax releases H3-Image weights, says model can generate 4K images in under 10 seconds

    MiniMax has released H3-Image, an artificial intelligence model designed to generate images from text and edit existing images. The company claims the model can produce 4,096-by-4,096-pixel images in under 10 seconds and use up to 16 reference images per request. Its weights are available on Hugging Face, and the model is also offered through MiniMax's API.

    MiniMax's H3-Image model offers two distinct access points for potential users. By releasing the model's weights on Hugging Face, MiniMax allows developers and researchers to download and run the core AI model on their own computing infrastructure. This approach supports a deeper level of customization, local integration, and independent experimentation with the model's capabilities.

    Concurrently, MiniMax is making H3-Image available via its Application Programming Interface (API). This typically provides a more managed, service-oriented access method, where users can integrate the model's functions into their applications without needing to host or maintain the underlying software themselves. This dual strategy caters to different user needs, from those seeking granular control to those prioritizing ease of use and a managed service.

    The company states H3-Image can generate high-resolution images, specifically 4,096-by-4,096 pixels, in under 10 seconds, and process up to 16 reference images in a single request. These reported performance metrics suggest a focus on both processing speed and the ability to handle complex image generation or editing tasks requiring multiple contextual inputs.

    Share this article

    Go deeper
    Original reporting by Tech NodeWe don't republish, read the full story â†’

    Related reading

    6 stories