Model comparison
gemma-4-E4B-it-assistant vs tinygemma3_cifar
These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: gemma-4-E4B-it-assistant's KV cache at 32K is 1.8× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| gemma-4-E4B-it-assistant | tinygemma3_cifar | |
|---|---|---|
| Parameters | 79M | 39M |
| Architecture | gemma4-assistant | gemma3 |
| Layers | 4 | 8 |
| Native context | 131,072 | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 9 | 1 |
| Smallest quantization | 0.07 GiB | 0.04 GiB |
| Q4_K_M | 0.07 GiB | — |
| Licence | apache-2.0 | wtfpl |
KV cache by context
the term that decides long-context viability
| Context | gemma-4-E4B-it-assistant | tinygemma3_cifar | Ratio |
|---|---|---|---|
| 4,096 | 0.01 GiB | 0.06 GiB | 4.57× |
| 8,192 | 0.02 GiB | 0.08 GiB | 3.59× |
| 16,384 | 0.04 GiB | 0.09 GiB | 2.50× |
| 32,768 | 0.07 GiB | 0.12 GiB | 1.81× |
| 65,536 | 0.13 GiB | 0.19 GiB | 1.43× |
| 131,072 | 0.26 GiB | 0.31 GiB | 1.22× |