Model comparison
gemma4-E4B-it-abliterated vs Qwen3-Coder-30B-A3B-Instruct
These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma4-E4B-it-abliterated's KV cache at 32K is 5.9× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| gemma4-E4B-it-abliterated | Qwen3-Coder-30B-A3B-Instruct | |
|---|---|---|
| Parameters | 7.9B | 30.5B |
| Architecture | gemma4 | qwen3moe |
| Layers | 42 | 48 |
| Native context | 131,072 | 262,144 |
| Mixture of experts | no | yes, 128 experts |
| Quantizations published | 24 | 46 |
| Smallest quantization | 3.06 GiB | 7.46 GiB |
| Q4_K_M | — | 17.28 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | gemma4-E4B-it-abliterated | Qwen3-Coder-30B-A3B-Instruct | Ratio |
|---|---|---|---|
| 4,096 | 0.12 GiB | 0.38 GiB | 3.05× |
| 8,192 | 0.18 GiB | 0.75 GiB | 4.22× |
| 16,384 | 0.29 GiB | 1.50 GiB | 5.22× |
| 32,768 | 0.51 GiB | 3.00 GiB | 5.93× |
| 65,536 | 0.94 GiB | 6.00 GiB | 6.36× |
| 131,072 | 1.82 GiB | 12.00 GiB | 6.60× |