Model comparison
Qwen2.5-3B-Instruct vs gemma-4-E4B-it
At Q4_K_M, Qwen2.5-3B-Instruct is the smaller download — 1,929,902,496 bytes against 4,977,171,584. At long context the gap widens: gemma-4-E4B-it's KV cache at 32K is 2.2× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Qwen2.5-3B-Instruct | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 3.1B | 8.0B |
| Architecture | qwen2 | gemma4 |
| Layers | 36 | 42 |
| Native context | 32,768 | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 1 | 36 |
| Smallest quantization | 1.80 GiB | 3.30 GiB |
| Q4_K_M | 1.80 GiB | 4.64 GiB |
| Licence | mit | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Qwen2.5-3B-Instruct | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.14 GiB | 0.12 GiB | 1.14× |
| 8,192 | 0.28 GiB | 0.18 GiB | 1.58× |
| 16,384 | 0.56 GiB | 0.29 GiB | 1.96× |
| 32,768 | 1.13 GiB | 0.51 GiB | 2.22× |
| 65,536 | 2.25 GiB | 0.94 GiB | 2.39× |
| 131,072 | 4.50 GiB | 1.82 GiB | 2.47× |