Model comparison
Trinity-Nano-Preview vs Qwen3-VL-8B-Instruct-abliterated-v1
At Q4_K_M, Trinity-Nano-Preview is the smaller download — 3,786,957,088 bytes against 5,027,785,792.
From the file· summed bytes, KV per layer
Side by side
| Trinity-Nano-Preview | Qwen3-VL-8B-Instruct-abliterated-v1 | |
|---|---|---|
| Parameters | 6.1B | 8.8B |
| Architecture | afmoe | qwen3vl |
| Layers | 56 | — |
| Native context | 131,072 | — |
| Mixture of experts | yes, 128 experts | no |
| Quantizations published | 50 | 48 |
| Smallest quantization | 1.94 GiB | 1.97 GiB |
| Q4_K_M | 3.53 GiB | 4.68 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Trinity-Nano-Preview | Qwen3-VL-8B-Instruct-abliterated-v1 | Ratio |
|---|---|---|---|
| 4,096 | 0.16 GiB | — | — |
| 8,192 | 0.21 GiB | — | — |
| 16,384 | 0.32 GiB | — | — |
| 32,768 | 0.54 GiB | — | — |
| 65,536 | 0.98 GiB | — | — |
| 131,072 | 1.85 GiB | — | — |