Model comparison
Nanbeige4.2-3B-heretic vs Qwen3-4B
At Q4_K_M, Qwen3-4B is the smaller download — 2,497,280,256 bytes against 2,574,808,288. At long context the gap widens: Nanbeige4.2-3B-heretic's KV cache at 32K is 1.6× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Nanbeige4.2-3B-heretic | Qwen3-4B | |
|---|---|---|
| Parameters | 4.2B | 4.0B |
| Architecture | nanbeige | qwen3 |
| Layers | 22 | 36 |
| Native context | 262,144 | 40,960 |
| Mixture of experts | no | no |
| Quantizations published | 4 | 29 |
| Smallest quantization | 2.40 GiB | 1.01 GiB |
| Q4_K_M | 2.40 GiB | 2.33 GiB |
| Licence | apache-2.0 | — |
KV cache by context
the term that decides long-context viability
| Context | Nanbeige4.2-3B-heretic | Qwen3-4B | Ratio |
|---|---|---|---|
| 4,096 | 0.34 GiB | 0.56 GiB | 1.64× |
| 8,192 | 0.69 GiB | 1.13 GiB | 1.64× |
| 16,384 | 1.38 GiB | 2.25 GiB | 1.64× |
| 32,768 | 2.75 GiB | 4.50 GiB | 1.64× |
| 65,536 | 5.50 GiB | 9.00 GiB | 1.64× |
| 131,072 | 11.00 GiB | 18.00 GiB | 1.64× |