Model comparison
NVIDIA-Nemotron-3-Nano-4B-BF16 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 2,837,072,896.
From the file· summed bytes, KV per layer
Side by side
| NVIDIA-Nemotron-3-Nano-4B-BF16 | HyperCLOVAX-SEED-Text-Instruct-1.5B | |
|---|---|---|
| Parameters | 4.0B | 1.6B |
| Architecture | nemotron_h | llama |
| Layers | 42 | — |
| Native context | 262,144 | — |
| Mixture of experts | no | no |
| Quantizations published | 42 | 1 |
| Smallest quantization | 2.03 GiB | 1.06 GiB |
| Q4_K_M | 2.64 GiB | 1.06 GiB |
| Licence | other | other |
KV cache by context
the term that decides long-context viability
| Context | NVIDIA-Nemotron-3-Nano-4B-BF16 | HyperCLOVAX-SEED-Text-Instruct-1.5B | Ratio |
|---|---|---|---|
| 4,096 | 0.66 GiB | — | — |
| 8,192 | 1.31 GiB | — | — |
| 16,384 | 2.63 GiB | — | — |
| 32,768 | 5.25 GiB | — | — |
| 65,536 | 10.50 GiB | — | — |
| 131,072 | 21.00 GiB | — | — |