Model comparison

DeepSWE-Preview vs Qwen3.6-27B

At Q4_K_M, Qwen3.6-27B is the smaller download — 16,547,397,888 bytes against 19,762,149,536. At long context the gap widens: Qwen3.6-27B's KV cache at 32K is 4.0× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

DeepSWE-PreviewQwen3.6-27B
Parameters32.8B27.8B
Architectureqwen3qwen35
Layers6464
Native context40,960262,144
Mixture of expertsnono
Quantizations published3040
Smallest quantization8.40 GiB8.74 GiB
Q4_K_M18.40 GiB15.41 GiB
Licencemitapache-2.0

KV cache by context

the term that decides long-context viability
ContextDeepSWE-PreviewQwen3.6-27BRatio
4,0961.00 GiB0.25 GiB4.00×
8,1922.00 GiB0.50 GiB4.00×
16,3844.00 GiB1.00 GiB4.00×
32,7688.00 GiB2.00 GiB4.00×
65,53616.00 GiB4.00 GiB4.00×
131,07232.00 GiB8.00 GiB4.00×