Model comparison
gemma4-26b-fiction-bf16 vs Qwen3.8-27B
These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma4-26b-fiction-bf16's KV cache at 32K is 1.3× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| gemma4-26b-fiction-bf16 | Qwen3.8-27B | |
|---|---|---|
| Parameters | 25.8B | 27.8B |
| Architecture | gemma4 | qwen35 |
| Layers | 30 | 64 |
| Native context | 262,144 | 262,144 |
| Mixture of experts | yes, 128 experts | no |
| Quantizations published | 23 | 22 |
| Smallest quantization | 7.72 GiB | 8.39 GiB |
| Q4_K_M | — | 15.66 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | gemma4-26b-fiction-bf16 | Qwen3.8-27B | Ratio |
|---|---|---|---|
| 4,096 | 0.45 GiB | 0.25 GiB | 1.80× |
| 8,192 | 0.61 GiB | 0.50 GiB | 1.21× |
| 16,384 | 0.92 GiB | 1.00 GiB | 1.09× |
| 32,768 | 1.54 GiB | 2.00 GiB | 1.30× |
| 65,536 | 2.79 GiB | 4.00 GiB | 1.43× |
| 131,072 | 5.29 GiB | 8.00 GiB | 1.51× |