Model comparison
Qwen3-The-Xiaolong-Omega-Directive-22B-uncensored-abliterated vs gemma-4-E4B-it
At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 13,333,473,760.
From the file· summed bytes, KV per layer
Side by side
| Qwen3-The-Xiaolong-Omega-Directive-22B-uncensored-abliterated | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 22.0B | 8.0B |
| Architecture | qwen3 | gemma4 |
| Layers | — | 42 |
| Native context | — | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 45 | 36 |
| Smallest quantization | 4.75 GiB | 3.30 GiB |
| Q4_K_M | 12.42 GiB | 4.64 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Qwen3-The-Xiaolong-Omega-Directive-22B-uncensored-abliterated | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | — | 0.12 GiB | — |
| 8,192 | — | 0.18 GiB | — |
| 16,384 | — | 0.29 GiB | — |
| 32,768 | — | 0.51 GiB | — |
| 65,536 | — | 0.94 GiB | — |
| 131,072 | — | 1.82 GiB | — |