Model comparison
gemma-4-26B-A4B-it-DFlash vs embeddinggemma-300m
These two publish different quantization sets; the table below has the exact sizes.
From the file· summed bytes, KV per layer
Side by side
| gemma-4-26B-A4B-it-DFlash | embeddinggemma-300m | |
|---|---|---|
| Parameters | 430M | 303M |
| Architecture | dflash | gemma-embedding |
| Layers | 5 | 24 |
| Native context | 262,144 | 2,048 |
| Mixture of experts | no | no |
| Quantizations published | 18 | 10 |
| Smallest quantization | 0.16 GiB | 0.26 GiB |
| Q4_K_M | 0.25 GiB | — |
| Licence | — | — |
KV cache by context
the term that decides long-context viability
| Context | gemma-4-26B-A4B-it-DFlash | embeddinggemma-300m | Ratio |
|---|---|---|---|
| 4,096 | 0.05 GiB | 0.04 GiB | 1.56× |
| 8,192 | 0.07 GiB | 0.05 GiB | 1.38× |
| 16,384 | 0.10 GiB | 0.08 GiB | 1.24× |
| 32,768 | 0.16 GiB | 0.14 GiB | 1.14× |
| 65,536 | 0.29 GiB | 0.27 GiB | 1.07× |
| 131,072 | 0.54 GiB | 0.52 GiB | 1.04× |