Pyyan / Compare / Hy4 preview vs MiMo-V2.6-Flash
Hy4 preview vs MiMo-V2.6-Flash
2 of 5Open-Weight Models · verified 29 Sept 2026
Hy4 previewTencentcurrentXMiMo-V2.6-FlashXiaomicurrent
| Specification | Hy4 preview | MiMo-V2.6-Flash |
|---|---|---|
| Summary | 770B parameters under Apache 2.0, and it helped optimise its own training. | The small one, and still multimodal. |
| Context | 1M | 1M |
| Input ($/Mtok) | $0.834 | Weights only |
| Output ($/Mtok) | $2.501 | Weights only |
| Licence | Apache 2.0 | MIT |
| Max output | 64K | Not published |
| Parameters | 770B total, 49B active | 309B total, 15B active |
| Released | 28 Aug 2026 | 21 Sep 2026 |
| Model ID | tencent/Hy4-preview | XiaomiMiMo/MiMo-V2.6-Flash |
| Category | Open-Weight Models | Open-Weight Models |
| Official | — | Xiaomi ↗ |
Highlighted rows are where these differ.
Hy4 preview
- 770B total parameters with 49B active, 78 layers, 256 routed experts plus one shared
- Apache 2.0, with an FP8 quantised build shipped alongside it
- Tencent used it during its own development, and report a 31.8% inference throughput gain from it
Best for the largest open weights.
MiMo-V2.6-Flash
- 309B total parameters, 15B active per token
- Same 1M context and five layer speculative decoder as Pro
- MIT licence, weights on Hugging Face
Best for multimodal work on your own cluster.