Model comparison
orpheus-3b-0.1-pretrained vs s2-pro
At Q4_K_M, orpheus-3b-0.1-pretrained is the smaller download — 2,363,760,640 bytes against 3,566,165,088.
From the file· summed bytes, KV per layer
Side by side
| orpheus-3b-0.1-pretrained | s2-pro | |
|---|---|---|
| Parameters | 3.8B | 4.6B |
| Architecture | llama | fish-speech |
| Layers | 28 | 36 |
| Native context | 131,072 | — |
| Mixture of experts | no | yes, 1 experts |
| Quantizations published | 45 | 7 |
| Smallest quantization | 1.40 GiB | 2.40 GiB |
| Q4_K_M | 2.20 GiB | 3.32 GiB |
| Licence | apache-2.0 | other |
KV cache by context
the term that decides long-context viability
| Context | orpheus-3b-0.1-pretrained | s2-pro | Ratio |
|---|---|---|---|
| 4,096 | 0.44 GiB | — | — |
| 8,192 | 0.88 GiB | — | — |
| 16,384 | 1.75 GiB | — | — |
| 32,768 | 3.50 GiB | — | — |
| 65,536 | 7.00 GiB | — | — |
| 131,072 | 14.00 GiB | — | — |