Model comparison
gpt2 vs Qwen3.6-35B-A3B-DFlash
At Q4_K_M, gpt2 is the smaller download — 112,858,624 bytes against 235,691,744.
From the file· summed bytes, KV per layer
Side by side
| gpt2 | Qwen3.6-35B-A3B-DFlash | |
|---|---|---|
| Parameters | 137M | 386M |
| Architecture | gpt2 | dflash |
| Layers | 12 | 6 |
| Native context | — | 262,144 |
| Mixture of experts | no | no |
| Quantizations published | 49 | 18 |
| Smallest quantization | 0.06 GiB | 0.14 GiB |
| Q4_K_M | 0.11 GiB | 0.22 GiB |
| Licence | — | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | gpt2 | Qwen3.6-35B-A3B-DFlash | Ratio |
|---|---|---|---|
| 4,096 | — | 0.09 GiB | — |
| 8,192 | — | 0.12 GiB | — |
| 16,384 | — | 0.15 GiB | — |
| 32,768 | — | 0.21 GiB | — |
| 65,536 | — | 0.34 GiB | — |
| 131,072 | — | 0.59 GiB | — |