Model comparison
Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-Uncensored vs gemma-4-E4B-it
At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 20,414,848,064. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 8.9× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-Uncensored | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 33.6B | 8.0B |
| Architecture | qwen3moe | gemma4 |
| Layers | 36 | 42 |
| Native context | 262,144 | 131,072 |
| Mixture of experts | yes, 12 experts | no |
| Quantizations published | 39 | 36 |
| Smallest quantization | 6.50 GiB | 3.30 GiB |
| Q4_K_M | 19.01 GiB | 4.64 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-Uncensored | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.56 GiB | 0.12 GiB | 4.57× |
| 8,192 | 1.13 GiB | 0.18 GiB | 6.33× |
| 16,384 | 2.25 GiB | 0.29 GiB | 7.84× |
| 32,768 | 4.50 GiB | 0.51 GiB | 8.90× |
| 65,536 | 9.00 GiB | 0.94 GiB | 9.54× |
| 131,072 | 18.00 GiB | 1.82 GiB | 9.90× |