Model comparison
SPARK.Chroma_preview vs Krea-2-Turbo
These two publish different quantization sets; the table below has the exact sizes.
From the file· summed bytes, KV per layer
Side by side
| SPARK.Chroma_preview | Krea-2-Turbo | |
|---|---|---|
| Parameters | 8.9B | 12.8B |
| Architecture | — | qwen_image |
| Layers | — | — |
| Native context | — | — |
| Mixture of experts | no | no |
| Quantizations published | 5 | 33 |
| Smallest quantization | 4.71 GiB | 2.78 GiB |
| Q4_K_M | — | 6.72 GiB |
| Licence | apache-2.0 | other |
KV cache by context
the term that decides long-context viability
| Context | SPARK.Chroma_preview | Krea-2-Turbo | Ratio |
|---|---|---|---|
| 4,096 | — | — | — |
| 8,192 | — | — | — |
| 16,384 | — | — | — |
| 32,768 | — | — | — |
| 65,536 | — | — | — |
| 131,072 | — | — | — |