DaisyRose-27B vs Hy4 flash Base
Compare documented capabilities and access. A dash means a comparable figure is unavailable.
| Specification | ||
|---|---|---|
| Task | Agentic coding | Multimodal pretraining |
| Published capabilities | Post-trained from Qwen3.8-27B on 1,200 agent tasks. RadixArk reports 12% pass@1 on Terminal-Bench 4.1. | Pretrained checkpoint for continued training and adaptation. The Flash architecture uses a sparse Engram memory module. |
| Model size | 27B parameters | Approximately 220B total / 12B active per token |
| Access | Public BF16, FP8 and NVFP4 checkpoints | Downloadable under the Tencent Hy License |
| Sources | Publisher documentation | Model weights |
| Inputs | Text, Images | Text, Images |
| Context window | 262,144 tokens | — |
| Weights | Open weights · Apache-2.0 | Open weights · Tencent Hy License |
| Released | Nov 18, 2026 | Nov 18, 2026 |