Hy4 flash Base vs Aion 1.0 Instruct 2.3B
Compare documented capabilities and access. A dash means a comparable figure is unavailable.
| Specification | ||
|---|---|---|
| Task | Multimodal pretraining | On-device language generation |
| Published capabilities | Pretrained checkpoint for continued training and adaptation. The Flash architecture uses a sparse Engram memory module. | Windows-oriented instruction model for local text generation, including document and email tasks. |
| Model size | Approximately 220B total / 12B active per token | 2.3B parameters |
| Access | Downloadable under the Tencent Hy License | Downloadable checkpoint |
| Operating limits | — | 32K-token context window |
| Sources | Model weights | Model weights |
| Inputs | Text, Images | Text |
| Weights | Open weights · Tencent Hy License | Open weights · MIT |
| Released | Nov 18, 2026 | Nov 18, 2026 |