Today ainotis Join

My notis

ModelsPublished All news from that day

MiniMax released the weights of H3-Image, a 33.1B image model, under a community licence

MiniMax says H3-Image makes a 4K image in under 10 seconds; its card says separate written authorization under Section IV of the licence is requested through an application form.

By ai notis. An AI newsroom with a human editor. Every fact links to its source. How we work

Share

Check our sources · 9 facts from 2 sources

MiniMax released the weights of H3-Image, an image generation and editing model with 33.1 billion parameters in its diffusion transformer, according to its Hugging Face model card. MiniMax's blog post is dated 11 November.

The model takes text prompts and up to 16 reference images, and produces images of about 16 megapixels, from 4096x4096 up to 6272x2688. It uses six sampling steps with no classifier-free guidance. The repository is 68.65 GB in BF16, and the text encoder is a separate 66.7 GB download.

The weights come under the MiniMax-H3-Image Community License Agreement, and the card says a separate written authorization under Section IV of the licence is requested through an application form. The card gives no benchmark tables and no price for the API on MiniMax's platform.

MiniMax claims a 4K image in under 10 seconds and prompt adherence it calls state-of-the-art among open-weight models. These are MiniMax's own claims, and the post names no benchmark.

Share this story

Your reaction

We count reactions per story and day, never who reacted. The counts help us choose what goes in the monthly issue. If you are signed in, your own page shows yours too.

Check our sources

Every sentence above is checked against these 2 sources.

1 MiniMaxAI/MiniMax-H3-Image model cardMiniMax · 6 facts Open the source archived copy

  1. The MiniMax-H3-Image model card describes the model as "a state-of-the-art open-source model for image generation and editing", handling text-to-image and editing of one or more reference images, with up to 16 references.

    MiniMax-H3-Image is a state-of-the-art open-source model for image generation and editing from the MiniMax H3 family.
    cite

  2. The diffusion transformer has 33.1B parameters; the full repository is 68.65 GB in BF16, and the text encoder (66.7 GB) is downloaded separately from MiniMax-H3.

    Diffusion transformer: 33.1B parameters, a dense single-stream transformer with 50 layers
    cite

  3. The model outputs images of about 16 megapixels, from 4096x4096 (1:1) to 6272x2688 (21:9), using 6 sampling steps with no classifier-free guidance.

    It generates images of about 16 megapixels at every aspect ratio, from 4096x4096 (1:1) to 6272x2688 (21:9), directly and without an upscaler
    cite

  4. "MiniMax-H3-Image is released under the MiniMax-H3-Image Community License Agreement." The card says the separate written authorization described in Section IV of the licence is requested through an application form.

    MiniMax-H3-Image is released under the MiniMax-H3-Image Community License Agreement.
    cite

  5. The model card lists no benchmark tables, no API price and no release date.

    (absence) Card read in full: it carries no benchmark table, no API price and no release date.
    cite

  6. The model is available through the MiniMax platform (platform.minimax.io, or platform.minimax.cn in China), with endpoints for text-to-image, editing and prompt expansion.

    Use MiniMax-H3-Image directly via API. - Global: platform.minimax.io | CN: platform.minimax.cn ... Text-to-image ... Image editing ... Prompt expansion
    cite
2 MiniMax H3-Image: Bringing 4K Image Generation into the 10-Second EraMiniMax · 11 Nov 2026 · 3 facts Open the source archived copy

  1. MiniMax's blog post on H3-Image is dated 2026-11-11 and describes an open-weight image generation model supporting text-to-image, reference-based generation and image editing with up to 16 reference images.

    MiniMax H3-Image: Bringing 4K Image Generation into the 10-Second Era 2026-11-11 Today, we're releasing MiniMax H3-Image, an open-weight image generation model built on H3.
    cite

  2. MiniMax says H3-Image generates a 4K (4096 x 4096) image in under 10 seconds.

    H3-Image can generate a 4K (4096 � 4096) image in under 10 seconds
    cite

  3. MiniMax says H3-Image "delivers state-of-the-art prompt adherence among open-weight models"; the post gives no named benchmark or numerical score for it.

    H3-Image delivers state-of-the-art prompt adherence among open-weight models, with support for up to 16 reference images.
    cite

Topics

The morning email

On the mornings we publish: the three top stories and up to four short ones. Free.

We email you a link to confirm. An issue may include one sponsor, always labelled Sponsored · Advertisement. Our emails count opens and clicks, not who made them. Unsubscribe in one click. What we keep