Model comparison
Nanbeige4.2-3B vs gemma-4-E4B-it
At Q4_K_M, Nanbeige4.2-3B is the smaller download — 2,574,807,904 bytes against 4,977,171,584. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 5.4× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Nanbeige4.2-3B | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 4.2B | 8.0B |
| Architecture | nanbeige | gemma4 |
| Layers | 22 | 42 |
| Native context | 262,144 | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 22 | 36 |
| Smallest quantization | 1.64 GiB | 3.30 GiB |
| Q4_K_M | 2.40 GiB | 4.64 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Nanbeige4.2-3B | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.34 GiB | 0.12 GiB | 2.79× |
| 8,192 | 0.69 GiB | 0.18 GiB | 3.87× |
| 16,384 | 1.38 GiB | 0.29 GiB | 4.79× |
| 32,768 | 2.75 GiB | 0.51 GiB | 5.44× |
| 65,536 | 5.50 GiB | 0.94 GiB | 5.83× |
| 131,072 | 11.00 GiB | 1.82 GiB | 6.05× |