Hunyuan-Large
Hunyuan-Large is the earlier downloadable Hunyuan-MoE-A52B family. It has 389B total parameters and uses 52B per token.
Model design and access
The release describes synthetic-data training and expert routing for language tasks. Its context limits differ between pretraining and instruction tuning. It remains useful as an identifiable earlier release, but it should not be presented as the current Hy flagship or assigned Hy3’s Apache license.Hunyuan-Large official model card
Practical use
Practical guidance. Use it when reproducing prior results or maintaining a deployment built for this architecture. Compare its infrastructure needs with the current releases before starting a new service.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| License | Tencent Hunyuan custom community licenseHunyuan-Large official model card |
| Context by training stage | Pretrained: 256K; instruction-tuned: 128KHunyuan-Large official model card |
Related models
Hy3 is a 295B mixture-of-experts language model with 21B active parameters and a separate 3.8B multi-token prediction layer.
Launch date not confirmedDownloadable · API- Hunyuan-A13B Instruct
Tencent · Hunyuan-A13B
An instruction-tuned mixture-of-experts language model with 80B total and 13B active parameters.
Launch date not confirmedDownloadable