Model comparison
Nanbeige4.2-3B-heretic 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 2,574,808,288.
From the file· summed bytes, KV per layer
Side by side
| Nanbeige4.2-3B-heretic | HyperCLOVAX-SEED-Text-Instruct-1.5B | |
|---|---|---|
| Parameters | 4.2B | 1.6B |
| Architecture | nanbeige | llama |
| Layers | 22 | — |
| Native context | 262,144 | — |
| Mixture of experts | no | no |
| Quantizations published | 4 | 1 |
| Smallest quantization | 2.40 GiB | 1.06 GiB |
| Q4_K_M | 2.40 GiB | 1.06 GiB |
| Licence | apache-2.0 | other |
KV cache by context
the term that decides long-context viability
| Context | Nanbeige4.2-3B-heretic | HyperCLOVAX-SEED-Text-Instruct-1.5B | Ratio |
|---|---|---|---|
| 4,096 | 0.34 GiB | — | — |
| 8,192 | 0.69 GiB | — | — |
| 16,384 | 1.38 GiB | — | — |
| 32,768 | 2.75 GiB | — | — |
| 65,536 | 5.50 GiB | — | — |
| 131,072 | 11.00 GiB | — | — |