Model comparison
Qwen3-Coder-30B-A3B-Instruct vs DeepSeek-V2-Lite-Chat-Uncensored
At Q4_K_M, DeepSeek-V2-Lite-Chat-Uncensored is the smaller download — 10,357,571,552 bytes against 18,556,689,568. At long context the gap widens: DeepSeek-V2-Lite-Chat-Uncensored's KV cache at 32K is 3.2× 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 | DeepSeek-V2-Lite-Chat-Uncensored | |
|---|---|---|
| Parameters | 30.5B | 15.7B |
| Architecture | qwen3moe | deepseek2 |
| Layers | 48 | 27 |
| Native context | 262,144 | 163,840 |
| Mixture of experts | yes, 128 experts | yes, 64 experts |
| Quantizations published | 46 | 11 |
| Smallest quantization | 7.46 GiB | 5.98 GiB |
| Q4_K_M | 17.28 GiB | 9.65 GiB |
| Licence | apache-2.0 | other |
KV cache by context
the term that decides long-context viability
| Context | Qwen3-Coder-30B-A3B-Instruct | DeepSeek-V2-Lite-Chat-Uncensored | Ratio |
|---|---|---|---|
| 4,096 | 0.38 GiB | 0.12 GiB | 3.16× |
| 8,192 | 0.75 GiB | 0.24 GiB | 3.16× |
| 16,384 | 1.50 GiB | 0.47 GiB | 3.16× |
| 32,768 | 3.00 GiB | 0.95 GiB | 3.16× |
| 65,536 | 6.00 GiB | 1.90 GiB | 3.16× |
| 131,072 | 12.00 GiB | 3.80 GiB | 3.16× |