image generation
HunyuanImage 3.0
Tencent open-weight autoregressive MoE image model under the Hunyuan Community licence, which excludes the EU, UK and South Korea.
Levelmodel
Classimage generation
Parameters80B
Contextnot in the notes
Released28 Sep 2025
Opennessopen weights restricted
Runs onmulti-GPU node
Statusactive
Benchmark Results
independent
Artificial Analysis Text-to-Image Arena AA-Image-T2I v2.0generative media
approximate rankindependent
945
5009001,300
Score in Elo; the axis starts at 500, as Elo has no zero. Higher is better.
rank 80, AA Text-to-Image Arena, overall
Where this comes from
Every score names who ran it: vendor-reported, maintainer-run or independent. A rank is shown with the population it was ranked in (gap rule 3).
Lineage
instruct ofsuccessor
HunyuanImage 3.0 Instruct is a child (instruct of)medium confidencenotesHunyuanImage 3.5 Preview is a child (successor)medium confidencenotesWhere this comes from
Parent and child relations from model cards, Hugging Face base_model fields and the genealogy tables of the research notes. The tree shows two generations each way, 3 models. Confidence and evidence are the notes' own.
R&D Topics
Identity
- Developer
- Tencent
- Family
- HunyuanImage
- Released
- 28 Sep 2025 sourcessecondary report
- Weights published
- –
- Record checked
- 10 Oct 2026
Where this comes from
Developer and dates from the research notes. Release date is the vendor's public announcement; weights date is when open weights were published.
Architecture and Size
- Architecture
- MoE Autoregressive MoE plus Transfusion, 64 experts
- Total parameters
- 80B sources
- Active parameters
- 13B
- Checkpoint parameters
- –
- Context length
- not in the notes
- Input
- text
- Output
- image
Where this comes from
Parameter counts as the developer states them. Where the checkpoint holds a different total, both are shown (register A16).
Openness and Access
- Openness
- open weights restricted
- Licence
- Tencent Hunyuan Community License sources
- Access routes
- open weights download
Where this comes from
The licence file governs over registry tags and aggregator labels.
Deployment
- Runs on
- multi-GPU node
- Footprint
- 3 to 4 x 80 GB GPUs
- Runtimes
- not in the notes
Where this comes from
Deployment tiers and runtimes as named in the notes. Footprint is the notes' figure, not a measurement by AURION.
Indonesia
- Indonesian support
- not stated
- Regional languages
- not in the notes
- Jakarta hosting
- unknown
- Languages
- not in the notes
Where this comes from
Indonesia fields are carried on every model; unknown is an allowed value.
Industry Use Cases
No industry use case links this model.
Links
DeveloperTencentbig tech, China
Tencentbig tech, ChinaModelHunyuanImage 3.0Tencent, image generation
HunyuanImage 3.0Tencent, image generationR&D TopicsUnified Multimodal Models for Understanding and GenerationText-to-Image Foundation Models (Diffusion and Flow Transformers)
Use Casesnone recorded
Who Runs Itnone recorded
Models 3
Organisations 1
R&D topics 2
Sources
- Tencent Open Sources Hunyuan Image 3.0 - World's Largest Open-Source Text-to-Image Model | ComfyUI Wiki comfyui-wiki.com https://comfyui-wiki.com/news/2025-09-27-tencent-open-source-hunyuan-image-3-0
- AA-Image-T2I v2.0 Leaderboard - Top AI Image Models | Artificial Analysis artificialanalysis.ai https://artificialanalysis.ai/text-to-image/arena/leaderboard-text
- Who we are tencent.com https://www.tencent.com/en-us/about.html
To cite this page: Project AURION, AI Use-Cases Research & Innovation, "HunyuanImage 3.0", https://aurion.semai.id/models/tencent-hunyuanimage-3-0/, records of October 2026.