Model comparison
Delphi-25B-SimpleRL-Math vs gemma-4-E4B-it
At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 15,160,698,624. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 66.2× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Delphi-25B-SimpleRL-Math | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 25.0B | 8.0B |
| Architecture | qwen3 | gemma4 |
| Layers | 51 | 42 |
| Native context | 16,384 | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 34 | 36 |
| Smallest quantization | 5.31 GiB | 3.30 GiB |
| Q4_K_M | 14.12 GiB | 4.64 GiB |
| Licence | — | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Delphi-25B-SimpleRL-Math | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 4.18 GiB | 0.12 GiB | 34.00× |
| 8,192 | 8.37 GiB | 0.18 GiB | 47.08× |
| 16,384 | 16.73 GiB | 0.29 GiB | 58.29× |
| 32,768 | 33.47 GiB | 0.51 GiB | 66.16× |
| 65,536 | 66.94 GiB | 0.94 GiB | 70.96× |
| 131,072 | 133.88 GiB | 1.82 GiB | 73.62× |