Model comparison
Marco-Nano-Instruct 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 5,371,989,536.
From the file· summed bytes, KV per layer
Side by side
| Marco-Nano-Instruct | HyperCLOVAX-SEED-Text-Instruct-1.5B | |
|---|---|---|
| Parameters | 8.0B | 1.6B |
| Architecture | qwen3moe | llama |
| Layers | 28 | — |
| Native context | 32,768 | — |
| Mixture of experts | yes, 232 experts | no |
| Quantizations published | 36 | 1 |
| Smallest quantization | 2.44 GiB | 1.06 GiB |
| Q4_K_M | 5.00 GiB | 1.06 GiB |
| Licence | apache-2.0 | other |
KV cache by context
the term that decides long-context viability
| Context | Marco-Nano-Instruct | HyperCLOVAX-SEED-Text-Instruct-1.5B | 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 | — | — |