What it does
The Tencent-Hunyuan GitHub organization collects public repositories for the Hunyuan family of models and related research code, with 88 repositories listed. Each model lives in its own repository rather than in a single app. The pinned projects cover several kinds of model. Hy3 is described as a reasoning and agent model listed as 295B A21B. HunyuanVideo and HunyuanVideo-I2V are for video generation, the latter from an image. HunyuanImage-3.0 is a multimodal model for image generation, Hunyuan3D-2.1 turns images into 3D assets with PBR materials, and HunyuanOCR is a lightweight OCR vision-language model. Other repositories include AuK for speech generation and editing, HunyuanDiT, a diffusion transformer with fine-grained Chinese understanding, the UniRL reinforcement learning framework and several benchmarks. The organization page links to hy.tencent.com as its homepage.
Capabilities
- Hy3, a reasoning and agent model
- HunyuanVideo and HunyuanVideo-I2V for video generation
- HunyuanImage-3.0 for image generation
- Hunyuan3D-2.1 for turning images into 3D assets
- HunyuanOCR for lightweight OCR with a vision-language model
- AuK for speech generation and editing, plus benchmarks and training frameworks
When to use it
- Looking for publicly hosted Tencent model code for text, image, video, 3D or speech work
- Comparing or benchmarking multimodal models and their training frameworks
- Following new releases from the Hunyuan team
