Hy4 Base vs Hy4 flash Base
Compare documented capabilities and access. A dash means a comparable figure is unavailable.
| Specification | ||
|---|---|---|
| Task | Multimodal pretraining | Multimodal pretraining |
| Published capabilities | Pretrained checkpoint for continued training and adaptation. Tencent’s flagship Hy4 uses a 78-layer language backbone with a separate vision encoder. | Pretrained checkpoint for continued training and adaptation. The Flash architecture uses a sparse Engram memory module. |
| Model size | 770B total / 49B active per token in the language backbone | Approximately 220B total / 12B active per token |
| Access | Downloadable under the Tencent Hy License | Downloadable under the Tencent Hy License |
| Sources | Model weights | Model weights |
| Inputs | Text, Images, Video | Text, Images |
| Weights | Open weights · Tencent Hy License | Open weights · Tencent Hy License |
| Released | Nov 20, 2026 | Nov 18, 2026 |