Model comparison
umt5-xxl vs gemma-4-E4B-it
At Q4_K_M, umt5-xxl is the smaller download — 3,655,145,312 bytes against 4,977,171,584.
From the file· summed bytes, KV per layer
Side by side
| umt5-xxl | gemma-4-E4B-it | |
|---|---|---|
| Parameters | 5.7B | 8.0B |
| Architecture | t5encoder | gemma4 |
| Layers | — | 42 |
| Native context | — | 131,072 |
| Mixture of experts | no | no |
| Quantizations published | 28 | 36 |
| Smallest quantization | 2.36 GiB | 3.30 GiB |
| Q4_K_M | 3.40 GiB | 4.64 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | umt5-xxl | gemma-4-E4B-it | Ratio |
|---|---|---|---|
| 4,096 | — | 0.12 GiB | — |
| 8,192 | — | 0.18 GiB | — |
| 16,384 | — | 0.29 GiB | — |
| 32,768 | — | 0.51 GiB | — |
| 65,536 | — | 0.94 GiB | — |
| 131,072 | — | 1.82 GiB | — |