MiniMax H3-Image
MiniMax's 33B open image model makes 16-megapixel pictures in six steps and edits with up to 16 references, with territory limits.
In short. MiniMax H3-Image is an AI image model you can download. It draws very large pictures, about 16 megapixels such as 4096 by 4096, in only six steps, and it can also edit pictures using up to 16 reference images. The license is not open in the usual sense: it excludes some regions and has a low income limit for commercial use.
Summary
MiniMax released H3-Image as open weights on 10 November 2026. It comes from the MiniMax H3 family, and one checkpoint handles both text-to-image generation and editing with one or more reference images. Its diffusion transformer has 33.1 billion parameters and is the H3 architecture without the audio layers; the text encoder is the Qwen3-VL-based encoder from MiniMax H3, unchanged. The checkpoint is distilled to 6 sampling steps with no classifier-free guidance.
The headline feature is resolution. The model writes images of about 16 megapixels directly, from 4096 x 4096 at 1:1 to 6272 x 2688 at 21:9, with no upscaler, and its stated working range is about 4 to 20 megapixels. It accepts up to 16 reference images and prompts of up to 4,096 tokens. The model card says it expects a detailed English description of the whole image, so for short prompts in any language MiniMax offers a hosted prompt-rewriting API and a local alternative, MiniMax-H3-Image-PE, an NVFP4 prompt expander of about 17.9 GB that is derived from Qwen3.8-27B. MiniMax says a 4K image takes under 10 seconds on its service, up to 5 times faster than mainstream alternatives, and calls the model's prompt-following the strongest among open-weight image models; these claims are its own, and no hardware is given.
The repository holds about 68.65 GB of BF16 transformer and autoencoder files, and the 66.7 GB text encoder is loaded from the MiniMax-H3 repository, which brings the full set to roughly 135 GB. SGLang, vLLM-Omni and ComfyUI can run it, on NVIDIA and AMD accelerators. The license is the catch. It covers only the world outside the European Union, the United Kingdom, South Korea and the United States, and it needs MiniMax's written authorization if you and your affiliates earn more than US$1 million a year or if you offer the model as a hosted service to anyone else.
Good for
- Very high-resolution images and large prints without a separate upscaling step
- Editing and compositing that mixes up to 16 reference images in one prompt
- Local image generation with only 6 sampling steps per image
- Individuals and small projects outside the European Union, the United Kingdom, South Korea and the United States
Watch out for
- The license does not cover the European Union, the United Kingdom, South Korea or the United States
- Authorization is required above US$1 million in yearly revenue, or if you run a hosted service for others, whatever your revenue
- The full set is roughly 135 GB, and MiniMax publishes no minimum GPU memory
- It expects long, detailed English prompts, so short prompts work best after a prompt-expansion step
- The license forbids using the model or its outputs to train other AI models
Performance claims come from the developer and have not been independently verified.
How much memory does it need?
Memory figures for each variant of MiniMax H3-Image. Image, video, speech and music pipelines also load a text encoder and a decoder, so the memory needed can be larger than the main checkpoint.
| Variants | Parameters | Download size | Memory needed | Fits on |
|---|---|---|---|---|
| H3-Image (BF16) BF16 | 33.1B | 135 GB | 155 GB | Big-memory desktop |
Download size covers every component needed to run the model, such as the generator, text encoder and decoder. Memory needed is a stated or published measured minimum where one exists; otherwise it is the download size plus about 15%. Larger images, longer clips and higher resolutions need more.
The license in plain words
Built on the MiniMax H3 license, with a much lower commercial threshold. You may use, modify and distribute MiniMax-H3-Image only in the "Applicable Territory" (the world except the European Union, the United Kingdom, South Korea and the United States). You need MiniMax's written authorization if you and your affiliates earn more than US$1 million a year, or if you offer the model as a hosted service to anyone else. Get legal advice before building a business on it.
- Attribution or model name must be shown
- Free commercial use only below a revenue cap
- Not licensed in some regions
- Acceptable-use policy applies
- Outputs must be filtered or reviewed, and AI generation disclosed
More about MiniMax H3-Image Community
See also
- Best open-source image and video generation models
- Free until you grow: where open licenses draw the line
- MiniMax H3-Image: 16-megapixel open weights, but not for the US or EU
- MOSS-Image: a 5B open image model for long text and JSON layouts
- Open licenses: free to build on, but hosting needs a deal
Frequently asked questions
- What is MiniMax H3-Image?
- MiniMax H3-Image is an AI image model you can download. It draws very large pictures, about 16 megapixels such as 4096 by 4096, in only six steps, and it can also edit pictures using up to 16 reference images. The license is not open in the usual sense: it excludes some regions and has a low income limit for commercial use.
- Can I use MiniMax H3-Image commercially?
- Yes, with conditions. MiniMax H3-Image is released under the MiniMax H3-Image Community, which allows commercial use but adds extra rules: Attribution or model name must be shown; Free commercial use only below a revenue cap; Not licensed in some regions; Acceptable-use policy applies; Outputs must be filtered or reviewed, and AI generation disclosed. Read the license text before you ship a product.
- Is MiniMax H3-Image open source?
- MiniMax H3-Image is open-weight, not open source in the strict sense. You can download the model files, but the MiniMax H3-Image Community adds conditions that the Open Source Initiative definition does not allow.
- Can I run MiniMax H3-Image on my own computer?
- Only on a machine with a lot of memory. The smallest variant, H3-Image (BF16), needs about 155 GB at 4-bit precision. Desktops with 128 to 512 GB of unified memory can hold it, at modest speed.
- When was MiniMax H3-Image released?
- MiniMax H3-Image was released by MiniMax on November 10, 2026.
Official links
Sources
Specifications come from these primary sources. Benchmark claims, where mentioned, are reported by the developer.
- MiniMax H3-Image: Bringing 4K Image Generation into the 10-Second Era (MiniMax) minimax.io
- MiniMax-H3-Image model card (Hugging Face) huggingface.co
- MiniMax-H3-Image Community License Agreement huggingface.co
- MiniMax H3 Image is now available in ComfyUI (ComfyUI blog) blog.comfy.org
Release date and facts last verified November 11, 2026.
Related models
| Model | Released | Parameters | Context | License | Smallest variant needs |
|---|---|---|---|---|---|
| FLUX 3 Image Klein Black Forest Labs | 5.6B | – | FLUX Kommunity (Conditional) | 8 GB Laptop or phone | |
| Qwen-Image-2.1 Alibaba Qwen | 7.1B | – | Qwen Research License (Non-commercial) | 20 GB Gaming GPU | |
| MiniMax H3 MiniMax | 33.1B | – | MiniMax H3 Community (Conditional) | 166 GB Big-memory desktop | |
| MOSS-Image OpenMOSS | 5.08B | – | Apache 2.0 (Permissive) | 4.2 GB Laptop or phone |