Model comparison
Mixtral-8x22B-v0.1 vs Qwen3-Coder-30B-A3B-Instruct
At Q4_K_M, Qwen3-Coder-30B-A3B-Instruct is the smaller download — 18,556,689,568 bytes against 85,586,760,928. At long context the gap widens: Qwen3-Coder-30B-A3B-Instruct's KV cache at 32K is 2.3× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Mixtral-8x22B-v0.1 | Qwen3-Coder-30B-A3B-Instruct | |
|---|---|---|
| Parameters | 141B | 30.5B |
| Architecture | llama | qwen3moe |
| Layers | 56 | 48 |
| Native context | 65,536 | 262,144 |
| Mixture of experts | yes, 8 experts | yes, 128 experts |
| Quantizations published | 28 | 46 |
| Smallest quantization | 16.99 GiB | 7.46 GiB |
| Q4_K_M | 79.71 GiB | 17.28 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Mixtral-8x22B-v0.1 | Qwen3-Coder-30B-A3B-Instruct | Ratio |
|---|---|---|---|
| 4,096 | 0.88 GiB | 0.38 GiB | 2.33× |
| 8,192 | 1.75 GiB | 0.75 GiB | 2.33× |
| 16,384 | 3.50 GiB | 1.50 GiB | 2.33× |
| 32,768 | 7.00 GiB | 3.00 GiB | 2.33× |
| 65,536 | 14.00 GiB | 6.00 GiB | 2.33× |
| 131,072 | 28.00 GiB | 12.00 GiB | 2.33× |