Hy4 flash Base vs Aurora2-400M-A20M
Compare documented capabilities and access. A dash means a comparable figure is unavailable.
| Specification | ||
|---|---|---|
| Task | Multimodal pretraining | Base language modeling |
| Published capabilities | Pretrained checkpoint for continued training and adaptation. The Flash architecture uses a sparse Engram memory module. | English base model pretrained on 100 billion tokens with one shared expert and 32 routed experts per layer. |
| Model size | Approximately 220B total / 12B active per token | 400M total / approximately 19M active parameters |
| Access | Downloadable under the Tencent Hy License | Public Transformers checkpoint for completion and adaptation |
| Sources | Model weights | Model weights |
| Inputs | Text, Images | Text |
| Weights | Open weights · Tencent Hy License | Open weights · Apache-2.0 |
| Released | Nov 18, 2026 | Nov 18, 2026 |