Model comparison
GigaChat3-10B-A1.8B-base vs Qwen3-30B-A3B-Thinking-2507
At Q4_K_M, GigaChat3-10B-A1.8B-base is the smaller download — 6,474,702,976 bytes against 18,556,685,824. At long context the gap widens: GigaChat3-10B-A1.8B-base's KV cache at 32K is 3.3× smaller, which usually matters more than the difference in weights.
From the file· summed bytes, KV per layer
Side by side
| GigaChat3-10B-A1.8B-base | Qwen3-30B-A3B-Thinking-2507 | |
|---|---|---|
| Parameters | 11.5B | 30.5B |
| Architecture | deepseek2 | qwen3moe |
| Layers | 26 | 48 |
| Native context | 262,144 | 262,144 |
| Mixture of experts | yes, 64 experts | yes, 128 experts |
| Quantizations published | 8 | 51 |
| Smallest quantization | 6.03 GiB | 7.05 GiB |
| Q4_K_M | 6.03 GiB | 17.28 GiB |
| Licence | mit | apache-2.0 |
KV cache by context
the term that decides long-context viability
| Context | GigaChat3-10B-A1.8B-base | Qwen3-30B-A3B-Thinking-2507 | Ratio |
|---|---|---|---|
| 4,096 | 0.11 GiB | 0.38 GiB | 3.28× |
| 8,192 | 0.23 GiB | 0.75 GiB | 3.28× |
| 16,384 | 0.46 GiB | 1.50 GiB | 3.28× |
| 32,768 | 0.91 GiB | 3.00 GiB | 3.28× |
| 65,536 | 1.83 GiB | 6.00 GiB | 3.28× |
| 131,072 | 3.66 GiB | 12.00 GiB | 3.28× |