Model comparison
pixtral-12b vs Qwythos-9B-Claude-Mythos-5-1M
At Q4_K_M, pixtral-12b is the smaller download — 7,477,204,608 bytes against 11,516,776,704. At long context the gap widens: Qwythos-9B-Claude-Mythos-5-1M's KV cache at 32K is 5.0× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| pixtral-12b | Qwythos-9B-Claude-Mythos-5-1M | |
|---|---|---|
| Parameters | 12.7B | 9.4B |
| Architecture | llama | qwen35 |
| Layers | 40 | 32 |
| Native context | 1,024,000 | 1,048,576 |
| Mixture of experts | no | no |
| Quantizations published | 32 | 16 |
| Smallest quantization | 3.85 GiB | 5.38 GiB |
| Q4_K_M | 6.96 GiB | 10.73 GiB |
| Licence | apache-2.0 | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | pixtral-12b | Qwythos-9B-Claude-Mythos-5-1M | Ratio |
|---|---|---|---|
| 4,096 | 0.63 GiB | 0.13 GiB | 5.00× |
| 8,192 | 1.25 GiB | 0.25 GiB | 5.00× |
| 16,384 | 2.50 GiB | 0.50 GiB | 5.00× |
| 32,768 | 5.00 GiB | 1.00 GiB | 5.00× |
| 65,536 | 10.00 GiB | 2.00 GiB | 5.00× |
| 131,072 | 20.00 GiB | 4.00 GiB | 5.00× |