Model comparison
rfdetr vs Supra-Router-51M
These two publish different quantization sets; the table below has the exact sizes.
From the file· summed bytes, KV per layer
Side by side
| rfdetr | Supra-Router-51M | |
|---|---|---|
| Parameters | 30M | 52M |
| Architecture | rfdetr | llama |
| Layers | — | 12 |
| Native context | — | 5,120 |
| Mixture of experts | no | no |
| Quantizations published | 12 | 18 |
| Smallest quantization | 0.03 GiB | 0.02 GiB |
| Q4_K_M | — | 0.03 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | rfdetr | Supra-Router-51M | Ratio |
|---|---|---|---|
| 4,096 | — | 0.05 GiB | — |
| 8,192 | — | 0.09 GiB | — |
| 16,384 | — | 0.19 GiB | — |
| 32,768 | — | 0.38 GiB | — |
| 65,536 | — | 0.75 GiB | — |
| 131,072 | — | 1.50 GiB | — |