MiniMax Open-Sources H3-Image Model for 10-Second 4K Generation
MiniMax has released H3-Image with open weights, an image generation model derived from its H3 base that supports text-to-image, image reference and image editing. It can produce a 4096x4096 (about 16MP) 4K image in 10 seconds, which MiniMax claims is roughly 5x faster than mainstream alternatives, and accepts up to 16 reference images with what it calls the strongest instruction following among open models.
Related Entities
Full Report
IT之家 November 10 news: late tonight (the 10th), MiniMax announced that MiniMax H3-Image is officially released and open-weights. This is an image generation model derived from H3, featuring text-to-image, image reference, and image editing capabilities.
According to the official introduction obtained by IT之家, it can generate a 4096 x 4096 pixel, approximately 16 MP (4K) high-resolution image within 10 seconds, while presenting rich, realistic details, and possessing excellent instruction-following and multi-reference-image understanding capabilities.
MiniMax states that, building on the broad image capabilities endowed by H3 pre-training, H3-Image further strengthens the entire post-training pipeline, doing a lot of solid work in areas such as data, reinforcement learning, step compression, PE, and deployment optimization. The combination of these efforts enables MiniMax H3-image to be competitive in three aspects simultaneously: speed, image quality, and instruction following:
- Compared with the mainstream 1-megapixel or 4-megapixel, a 16-megapixel (16MP, 4096×4096) image has a significant advantage in detail. MiniMax H3-Image generates a 16MP high-specification image within 10s, up to 5 times faster than mainstream levels on the market.
- Image details and realism can significantly broaden the model's range of use. Unlike the common "AI feel," H3-Image can provide rich image details and "extremely strong realism."
- The model provides the "strongest among open-source models" instruction-following capability, supporting up to 16 reference images.
During the first week of release, MiniMax will also offer a 50% discount on the official price for generation requests of 16MP and above.
- Hugging Face: https://huggingface.co/MiniMaxAI/MiniMax-H3-Image
- API: https://platform.minimax.cn/docs/guides/image-generation
- MiniMax Code: https://code.minimax.cn/about
Advertising statement: External redirect links contained in the article (including but not limited to hyperlinks, QR codes, passwords, and other forms) are used to convey more information and save selection time; the results are for reference only. All articles on IT之家 containing external links include this statement.