Model comparison
Qwen3.8-27B vs NVIDIA-Nemotron-Nano-12B-v2
At Q4_K_M, NVIDIA-Nemotron-Nano-12B-v2 is the smaller download — 7,494,497,504 bytes against 16,810,714,336. At long context the gap widens: Qwen3.8-27B's KV cache at 32K is 3.9× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Qwen3.8-27B | NVIDIA-Nemotron-Nano-12B-v2 | |
|---|---|---|
| Parameters | 27.8B | 12.3B |
| Architecture | qwen35 | nemotron_h |
| Layers | 64 | 62 |
| Native context | 262,144 | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 22 | 30 |
| Smallest quantization | 8.39 GiB | 3.79 GiB |
| Q4_K_M | 15.66 GiB | 6.98 GiB |
| Licence | apache-2.0 | — |
KV cache by context
the term that decides long-context viability
| Context | Qwen3.8-27B | NVIDIA-Nemotron-Nano-12B-v2 | Ratio |
|---|---|---|---|
| 4,096 | 0.25 GiB | 0.97 GiB | 3.87× |
| 8,192 | 0.50 GiB | 1.94 GiB | 3.87× |
| 16,384 | 1.00 GiB | 3.88 GiB | 3.87× |
| 32,768 | 2.00 GiB | 7.75 GiB | 3.87× |
| 65,536 | 4.00 GiB | 15.50 GiB | 3.87× |
| 131,072 | 8.00 GiB | 31.00 GiB | 3.87× |