Model comparison

oxy-1-small vs Qwen-AgentWorld-35B-A3B

These two publish different quantization sets; the table below has the exact sizes. At long context the gap widens: Qwen-AgentWorld-35B-A3B's KV cache at 32K is 9.6× smaller, which usually matters more than the difference in weights.

From the file· summed bytes, KV per layer

Side by side

oxy-1-smallQwen-AgentWorld-35B-A3B
Parameters14.8B34.7B
Architectureqwen2qwen35moe
Layers4840
Native context32,768262,144
Mixture of expertsnoyes, 256 experts
Quantizations published2215
Smallest quantization4.66 GiB10.71 GiB
Q4_K_M8.37 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
Contextoxy-1-smallQwen-AgentWorld-35B-A3BRatio
4,0960.75 GiB0.08 GiB9.59×
8,1921.50 GiB0.16 GiB9.60×
16,3843.00 GiB0.31 GiB9.60×
32,7686.00 GiB0.63 GiB9.60×
65,53612.00 GiB1.25 GiB9.60×
131,07224.00 GiB2.50 GiB9.60×