Event date · · MiniMax

Reportedly: MiniMax Releases H3-Image Open-Weight Model Generating 4K Images in 10 Seconds

MiniMax 稀宇科技Chinese AIOpen weights

Reported by IT Home · not yet confirmed by the company or a second independent outlet. We update this page when it is.

REPORTED

MiniMax released H3-Image, an open-weight image generation model derived from H3, capable of generating a 4096x4096 pixel (16MP) image in 10 seconds. It supports text-to-image, image reference, and image editing, with up to 16 reference images. The release was announced on November 10, 2026, IT Home reported.

China context

Original name
稀宇科技
Outside China
Open weights · ithome.com
Claims
Company-reported; not yet independently evaluated
For builders
Developers can access the model weights to self-host image generation, potentially avoiding per-image API costs and enabling customization.
For investors
The release of a high-speed, high-resolution open-weight model may intensify competition in the image generation market, affecting the value proposition of closed-source alternatives.

Translated from Chinese. Quotes and facts link to the original sources.

What happened

MiniMax announced the release and open-weight availability of H3-Image, an image generation model derived from its H3 model. The model can generate a 4096x4096 pixel (approximately 16MP) image in 10 seconds, which MiniMax claims is about 5 times faster than mainstream models. It offers text-to-image, image reference, and image editing capabilities, with support for up to 16 reference images. MiniMax states the model provides rich detail, strong realism, and the strongest instruction-following among open-source models.

Technical significance

H3-Image builds on H3's pretrained image capabilities and enhances the post-training pipeline, including data, reinforcement learning, step compression, and deployment optimization. The model generates 16MP images in 10 seconds, which MiniMax claims is 5 times faster than mainstream models. It supports up to 16 reference images for multi-reference understanding.

Industry impact

Developers outside China gain access to an open-weight model that can produce high-resolution 4K images in 10 seconds, potentially reducing the cost and time for image generation tasks compared to existing open models. This release may pressure other open-weight image model providers to improve speed and resolution.

Decision value

The model's open weights allow businesses to self-host image generation without per-image API costs, potentially lowering expenses for high-volume use cases. Its speed and resolution could enable new applications requiring rapid, high-quality image generation.

What to watch

Specific signals to watch include whether MiniMax releases the model weights on a public platform like Hugging Face, whether independent benchmarks confirm the claimed speed and quality, and whether developers adopt it for production use. The first-week promotion for 16MP and above specifications may indicate commercial API plans.

Latest in Chinese AI

  1. Ant GroupReportedly: Ant Group's Robbyant Open-Sources LingBot-Touch Tactile Model for Robot Hands
  2. HygonReportedly: Hygon Launches 14XX Industrial-Control CPU With 10W TDP and a -40°C to 105°C Operating Range
  3. Deep RoboticsReportedly: Deep Robotics releases Jueying X50 quadruped robot with 50kg payload and complex-terrain mobility
  4. Reportedly: Apple Intelligence generative AI service completes filing for iPad and MacBook in China
  5. ByteDanceReportedly: ByteDance releases Seedream 5.5 image model, integrated into Doubao, Doubao Work, and Jimeng

All China AI Events

AIGC Newsletter

China AI, with sources and context.

Analysis of Chinese AI models, companies and policy, and what you can use outside China.