Model comparison
moonshine-streaming-small vs whisper-large-v3-turbo
These two publish different quantization sets; the table below has the exact sizes.
From the file· summed bytes, KV per layer
Side by side
| moonshine-streaming-small | whisper-large-v3-turbo | |
|---|---|---|
| Parameters | 140M | 809M |
| Architecture | moonshine_streaming | whisper |
| Layers | 10 | 32 |
| Native context | 4,096 | — |
| Mixture of experts | no | no |
| Quantizations published | 3 | 31 |
| Smallest quantization | 0.18 GiB | 0.27 GiB |
| Q4_K_M | — | 0.50 GiB |
| Licence | mit | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | moonshine-streaming-small | whisper-large-v3-turbo | Ratio |
|---|---|---|---|
| 4,096 | 0.08 GiB | — | — |
| 8,192 | 0.16 GiB | — | — |
| 16,384 | 0.31 GiB | — | — |
| 32,768 | 0.63 GiB | — | — |
| 65,536 | 1.25 GiB | — | — |
| 131,072 | 2.50 GiB | — | — |