Model comparison

Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-Hightop vs Qwen3-Coder-30B-A3B-Instruct

At Q4_K_M, Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-Hightop is the smaller download — 7,323,108,608 bytes against 18,556,689,568. At long context the gap widens: Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-Hightop'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

Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-HightopQwen3-Coder-30B-A3B-Instruct
Parameters12.1B30.5B
Architectureqwen35qwen3moe
Layers2448
Native context262,144262,144
Mixture of expertsnoyes, 128 experts
Quantizations published3346
Smallest quantization4.60 GiB7.46 GiB
Q4_K_M6.82 GiB17.28 GiB
Licenceapache-2.0apache-2.0

KV cache by context

the term that decides long-context viability
ContextQwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-HightopQwen3-Coder-30B-A3B-InstructRatio
4,0960.09 GiB0.38 GiB4.00×
8,1920.19 GiB0.75 GiB4.00×
16,3840.38 GiB1.50 GiB4.00×
32,7680.75 GiB3.00 GiB4.00×
65,5361.50 GiB6.00 GiB4.00×
131,0723.00 GiB12.00 GiB4.00×
Qwen3.6-12B-IQ-Ultra-Heretic-Uncensored-Thinking-V2-Hightop vs Qwen3-Coder-30B-A3B-Instruct — size, memory and hardware fit — ossmodeldb