Model comparison
Qwen3-Coder-30B-A3B-Instruct vs Fallen-Gemma3-27B-v1
At Q4_K_M, Fallen-Gemma3-27B-v1 is the smaller download — 16,546,688,704 bytes against 18,556,689,568. At long context the gap widens: Fallen-Gemma3-27B-v1's KV cache at 32K is 1.3× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| Qwen3-Coder-30B-A3B-Instruct | Fallen-Gemma3-27B-v1 | |
|---|---|---|
| Parameters | 30.5B | 27.4B |
| Architecture | qwen3moe | gemma3 |
| Layers | 48 | 48 |
| Native context | 262,144 | — |
| Mixture of experts | yes, 128 experts | no |
| Quantizations published | 46 | 25 |
| Smallest quantization | 7.46 GiB | 7.86 GiB |
| Q4_K_M | 17.28 GiB | 15.41 GiB |
| Licence | apache-2.0 | other |
KV cache by context
the term that decides long-context viability
| Context | Qwen3-Coder-30B-A3B-Instruct | Fallen-Gemma3-27B-v1 | Ratio |
|---|---|---|---|
| 4,096 | 0.38 GiB | 0.67 GiB | 1.80× |
| 8,192 | 0.75 GiB | 0.91 GiB | 1.21× |
| 16,384 | 1.50 GiB | 1.38 GiB | 1.09× |
| 32,768 | 3.00 GiB | 2.31 GiB | 1.30× |
| 65,536 | 6.00 GiB | 4.19 GiB | 1.43× |
| 131,072 | 12.00 GiB | 7.94 GiB | 1.51× |