Model comparison
Mira-v1.12-Ties-27B vs Qwen3-Coder-30B-A3B-Instruct
These two publish different quantization sets; the table below has the exact sizes.
From the file· summed bytes, KV per layer
Side by side
| Mira-v1.12-Ties-27B | Qwen3-Coder-30B-A3B-Instruct | |
|---|---|---|
| Parameters | 27.4B | 30.5B |
| Architecture | gemma3 | qwen3moe |
| Layers | 62 | 48 |
| Native context | 131,072 | 262,144 |
| Mixture of experts | no | yes, 128 experts |
| Quantizations published | 23 | 46 |
| Smallest quantization | 5.83 GiB | 7.46 GiB |
| Q4_K_M | — | 17.28 GiB |
| Licence | — | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | Mira-v1.12-Ties-27B | Qwen3-Coder-30B-A3B-Instruct | Ratio |
|---|---|---|---|
| 4,096 | 0.92 GiB | 0.38 GiB | 2.46× |
| 8,192 | 1.23 GiB | 0.75 GiB | 1.65× |
| 16,384 | 1.86 GiB | 1.50 GiB | 1.24× |
| 32,768 | 3.11 GiB | 3.00 GiB | 1.04× |
| 65,536 | 5.61 GiB | 6.00 GiB | 1.07× |
| 131,072 | 10.61 GiB | 12.00 GiB | 1.13× |