Model comparison
gemma-4-26B-roleplay-v2-merged vs gpt-oss-20b
At Q4_K_M, gpt-oss-20b is the smaller download — 11,624,759,488 bytes against 16,796,015,616.
From the file· summed bytes, KV per layer
Side by side
| gemma-4-26B-roleplay-v2-merged | gpt-oss-20b | |
|---|---|---|
| Parameters | 25.2B | 21.5B |
| Architecture | gemma4 | gpt-oss |
| Layers | — | 24 |
| Native context | — | 131,072 |
| Mixture of experts | no | yes, 32 experts |
| Quantizations published | 34 | 15 |
| Smallest quantization | 7.72 GiB | 10.68 GiB |
| Q4_K_M | 15.64 GiB | 10.83 GiB |
| Licence | — | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | gemma-4-26B-roleplay-v2-merged | gpt-oss-20b | Ratio |
|---|---|---|---|
| 4,096 | — | 0.11 GiB | — |
| 8,192 | — | 0.21 GiB | — |
| 16,384 | — | 0.39 GiB | — |
| 32,768 | — | 0.77 GiB | — |
| 65,536 | — | 1.52 GiB | — |
| 131,072 | — | 3.02 GiB | — |