Model comparison
LFM2.5-8B-A1B-KO-SFT vs gemma-4-E4B-it
At Q4_K_M, gemma-4-E4B-it is the smaller download — 4,977,171,584 bytes against 5,155,565,504. At long context the gap widens: LFM2.5-8B-A1B-KO-SFT's KV cache at 32K is 1.3× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| LFM2.5-8B-A1B-KO-SFT | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 8.5B | 8.0B |
| Architecture | lfm2moe | gemma4 |
| Layers | 24 | 42 |
| Native context | 128,000 | 131,072 |
| Mixture of experts | yes, 32 experts | no |
| Quantizations published | 36 | 36 |
| Smallest quantization | 1.70 GiB | 3.30 GiB |
| Q4_K_M | 4.80 GiB | 4.64 GiB |
| Licence | other | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | LFM2.5-8B-A1B-KO-SFT | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | 0.05 GiB | 0.12 GiB | 2.62× |
| 8,192 | 0.09 GiB | 0.18 GiB | 1.90× |
| 16,384 | 0.19 GiB | 0.29 GiB | 1.53× |
| 32,768 | 0.38 GiB | 0.51 GiB | 1.35× |
| 65,536 | 0.75 GiB | 0.94 GiB | 1.26× |
| 131,072 | 1.50 GiB | 1.82 GiB | 1.21× |