Model comparison
gemma-4-12B-coder-fable5-composer2.5-v1-abliterated vs Qwen3-30B-A3B-Thinking-2507
At Q4_K_M, gemma-4-12B-coder-fable5-composer2.5-v1-abliterated is the smaller download — 8,568,894,912 bytes against 18,556,685,824. At long context the gap widens: gemma-4-12B-coder-fable5-composer2.5-v1-abliterated's KV cache at 32K is 1.2× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| gemma-4-12B-coder-fable5-composer2.5-v1-abliterated | Qwen3-30B-A3B-Thinking-2507 | |
|---|---|---|
| Parameters | 12.0B | 30.5B |
| Architecture | gemma4 | qwen3moe |
| Layers | 48 | 48 |
| Native context | 262,144 | 262,144 |
| Mixture of experts | no | yes, 128 experts |
| Quantizations published | 6 | 51 |
| Smallest quantization | 6.77 GiB | 7.05 GiB |
| Q4_K_M | 7.98 GiB | 17.28 GiB |
| Licence | gemma | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | gemma-4-12B-coder-fable5-composer2.5-v1-abliterated | Qwen3-30B-A3B-Thinking-2507 | Ratio |
|---|---|---|---|
| 4,096 | 0.72 GiB | 0.38 GiB | 1.92× |
| 8,192 | 0.97 GiB | 0.75 GiB | 1.29× |
| 16,384 | 1.47 GiB | 1.50 GiB | 1.02× |
| 32,768 | 2.47 GiB | 3.00 GiB | 1.22× |
| 65,536 | 4.47 GiB | 6.00 GiB | 1.34× |
| 131,072 | 8.47 GiB | 12.00 GiB | 1.42× |