Model comparison
Qwen3-Coder-30B-A3B-Instruct vs L3.2-8X4B-MOE-V2-Dark-Champion-Inst-21B-uncen-ablit
At Q4_K_M, L3.2-8X4B-MOE-V2-Dark-Champion-Inst-21B-uncen-ablit is the smaller download — 12,850,124,384 bytes against 18,556,689,568.
From the file· summed bytes, KV per layer
Side by side
| Qwen3-Coder-30B-A3B-Instruct | L3.2-8X4B-MOE-V2-Dark-Champion-Inst-21B-uncen-ablit | |
|---|---|---|
| Parameters | 30.5B | 20.9B |
| Architecture | qwen3moe | llama |
| Layers | 48 | — |
| Native context | 262,144 | — |
| Mixture of experts | yes, 128 experts | no |
| Quantizations published | 46 | 43 |
| Smallest quantization | 7.46 GiB | 4.35 GiB |
| Q4_K_M | 17.28 GiB | 11.97 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Qwen3-Coder-30B-A3B-Instruct | L3.2-8X4B-MOE-V2-Dark-Champion-Inst-21B-uncen-ablit | Ratio |
|---|---|---|---|
| 4,096 | 0.38 GiB | — | — |
| 8,192 | 0.75 GiB | — | — |
| 16,384 | 1.50 GiB | — | — |
| 32,768 | 3.00 GiB | — | — |
| 65,536 | 6.00 GiB | — | — |
| 131,072 | 12.00 GiB | — | — |