Model comparison
Amaretto-3B vs HyperCLOVAX-SEED-Text-Instruct-1.5B
At Q4_K_M, HyperCLOVAX-SEED-Text-Instruct-1.5B is the smaller download — 1,133,974,368 bytes against 2,146,498,240.
From the file· summed bytes, KV per layer
Side by side
| Amaretto-3B | HyperCLOVAX-SEED-Text-Instruct-1.5B | |
|---|---|---|
| Parameters | 4.3B | 1.6B |
| Architecture | mistral3 | llama |
| Layers | 26 | — |
| Native context | 262,144 | — |
| Mixture of experts | no | no |
| Quantizations published | 36 | 1 |
| Smallest quantization | 0.87 GiB | 1.06 GiB |
| Q4_K_M | 2.00 GiB | 1.06 GiB |
| Licence | apache-2.0 | other |
KV cache by context
the term that decides long-context viability
| Context | Amaretto-3B | HyperCLOVAX-SEED-Text-Instruct-1.5B | Ratio |
|---|---|---|---|
| 4,096 | 0.41 GiB | — | — |
| 8,192 | 0.81 GiB | — | — |
| 16,384 | 1.63 GiB | — | — |
| 32,768 | 3.25 GiB | — | — |
| 65,536 | 6.50 GiB | — | — |
| 131,072 | 13.00 GiB | — | — |