Model comparison
gemma-4-E4B-it-assistant vs LFM2.5-230M
At Q4_K_M, gemma-4-E4B-it-assistant is the smaller download — 77,231,200 bytes against 153,406,048. At long context the gap widens: gemma-4-E4B-it-assistant's KV cache at 32K is 5.5× 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 | LFM2.5-230M | |
|---|---|---|
| Parameters | 79M | 230M |
| Architecture | gemma4-assistant | lfm2 |
| Layers | 4 | 14 |
| Native context | 131,072 | 128,000 |
| Mixture of experts | no | no |
| Quantizations published | 9 | 25 |
| Smallest quantization | 0.07 GiB | 0.10 GiB |
| Q4_K_M | 0.07 GiB | 0.14 GiB |
| Licence | apache-2.0 | other |
KV cache by context
the term that decides long-context viability
| Context | gemma-4-E4B-it-assistant | LFM2.5-230M | Ratio |
|---|---|---|---|
| 4,096 | 0.01 GiB | 0.05 GiB | 3.43× |
| 8,192 | 0.02 GiB | 0.09 GiB | 4.36× |
| 16,384 | 0.04 GiB | 0.19 GiB | 5.05× |
| 32,768 | 0.07 GiB | 0.38 GiB | 5.49× |
| 65,536 | 0.13 GiB | 0.75 GiB | 5.73× |
| 131,072 | 0.26 GiB | 1.50 GiB | 5.86× |