MiniMax 发布开源图像生成模型 H3-Image

Today, we're releasing MiniMax H3-Image, an open-weight image generation model built on H3. It supports text-to-image generation, reference-based generation, and image editing.

H3-Image can generate a 4K (4096 × 4096) image in under 10 seconds, with rich, realistic details, strong prompt adherence, the ability to understand and work with up to 16 reference image.

Building on H3's pre-training capabilities, we've made improvements across the entire post-training pipeline, from data and reinforcement learning to sampling-step reduction, PE, and deployment optimization.

No single change made the difference. Together, these improvements allow H3-Image to deliver speed, image quality, and prompt adherence at the same time.

Speed: Making 4K Image Generation Practical for Everyday Use

Compared to the more common 1K or 2K outputs, 4K (16MP, 4096 × 4096) images offer significantly more detail.

MiniMax H3-Image can generate a full-resolution 4K image in under 10 seconds, at up to 5× the speed of mainstream alternatives.

This makes 4K image generation practical for everyday creative workflows.

Image Quality: Bringing AI-Generated Images Closer to Reality

H3-Image produces rich, natural details without the overly smooth, artificial look often associated with AI-generated images.From fine textures to subtle visual details, the model delivers a high degree of realism.

We want to further narrow the gap between AI-generated images and the real visual world.

Prompt Adherence: Understanding Creative Intent

High-quality image generation isn't just about producing beautiful images. It's also about accurately capturing and realizing the creator's intent.

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

Open Weights. More to Build.

As with H3, we're releasing H3-Image's model weights.

We want to support open research and give developers and creators the freedom to customize models and build their own tools and workflows.

We hope H3-Image can help narrow the gap between open-weight image models and leading proprietary models.

We look forward to seeing what developers, creators, and researchers build with it.

Availability and Pricing

H3-Image is also available through the MiniMax API Platform, MiniMax Code, and Hailuo AI, providing reliable, high-quality image generation services.

H3-Image's Standard mode is 1.5–2× as fast as mainstream alternatives, while Highspeed mode reaches 3–4× the speed.

Pricing is 50–70% of comparable mainstream offerings, making high-quality image generation more affordable.

For launch week, we're offering 50% off the standard price for image generation requests at 4K resolution and above.

Get started:

• Hugging Face: https://huggingface.co/MiniMaxAI/MiniMax-H3-Image
• API: https://platform.minimax.io/docs/guides/image-generation
• MiniMax Code: https://code.minimax.io/about

We hope H3-Image can be useful in more people's daily work and creative projects. We'll keep building better, more accessible tools to bring AGI to everyone.

Intelligence with Everyone.