HunyuanImage-3.0
HunyuanImage-3.0 uses an autoregressive multimodal design with 80B total and 13B active parameters. The original downloadable release generates images from text.
Model design and access
The release brings image and text processing into one model framework. The original checkpoint is distinct from the later Instruct model, which adds image-conditioned editing and instruction reasoning. The documented reference setup for text-to-image uses at least three 80 GB GPUs.HunyuanImage-3.0 official model documentation
Practical use
Practical guidance. Plan the serving system around the full weights, not the active parameter figure. Check layout, visual details and text accuracy on the same prompt set when comparing the original checkpoint with the Instruct release.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| License | Tencent Hunyuan Community License; excludes the EU, UK and South KoreaHunyuanImage 3.0 community license |
| Expert network | 64 experts; 13B active parametersHunyuanImage-3.0 official model documentation |
Related models
- Hy-Image-3.5 Preview
Tencent · Hy Image API
An image-generation and editing model with reference-image input and multi-turn editing.
Launch date not confirmedPreviewAPI - HunyuanImage-3.0 Instruct
Tencent · HunyuanImage
The Instruct release adds reasoning, prompt rewriting and image-to-image generation to HunyuanImage-3.0.
Launch date Downloadable