Model comparison
SuperHY3-abliterated-NVFP4 vs Hy3
These two publish different quantization sets; the table below has the exact sizes.
From the file· summed bytes, KV per layer
Side by side
| SuperHY3-abliterated-NVFP4 | Hy3 | |
|---|---|---|
| Parameters | 172B | 299B |
| Architecture | hy-v3 | hy_v3 |
| Layers | 80 | 80 |
| Native context | 262,144 | 262,144 |
| Mixture of experts | yes, 192 experts | yes, 192 experts |
| Quantizations published | 1 | 37 |
| Smallest quantization | 95.10 GiB | 59.65 GiB |
| Q4_K_M | — | 168.57 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | SuperHY3-abliterated-NVFP4 | Hy3 | Ratio |
|---|---|---|---|
| 4,096 | 1.25 GiB | 1.25 GiB | 1.00× |
| 8,192 | 2.50 GiB | 2.50 GiB | 1.00× |
| 16,384 | 5.00 GiB | 5.00 GiB | 1.00× |
| 32,768 | 10.00 GiB | 10.00 GiB | 1.00× |
| 65,536 | 20.00 GiB | 20.00 GiB | 1.00× |
| 131,072 | 40.00 GiB | 40.00 GiB | 1.00× |