Model comparison
umt5-xxl vs HyperCLOVAX-SEED-Text-Instruct-1.5B
At Q4_K_M, HyperCLOVAX-SEED-Text-Instruct-1.5B is the smaller download — 1,133,974,368 bytes against 3,655,145,312.
From the file· summed bytes, KV per layer
Side by side
| umt5-xxl | HyperCLOVAX-SEED-Text-Instruct-1.5B | |
|---|---|---|
| Parameters | 5.7B | 1.6B |
| Architecture | t5encoder | llama |
| Layers | — | — |
| Native context | — | — |
| Mixture of experts | no | no |
| Quantizations published | 28 | 1 |
| Smallest quantization | 2.36 GiB | 1.06 GiB |
| Q4_K_M | 3.40 GiB | 1.06 GiB |
| Licence | apache-2.0 | other |
KV cache by context
the term that decides long-context viability
| Context | umt5-xxl | HyperCLOVAX-SEED-Text-Instruct-1.5B | Ratio |
|---|---|---|---|
| 4,096 | — | — | — |
| 8,192 | — | — | — |
| 16,384 | — | — | — |
| 32,768 | — | — | — |
| 65,536 | — | — | — |
| 131,072 | — | — | — |