Qwen3.6-27B vs 35B, I prefer 35B but more people here post about 27B…

I've had better results quality wise with 35B AND it's much faster than 27B. Just curious cause I see lots of people post about 27B. Am I doing something wrong with 27B?

Use cases are multi-stage pipelines for coding and internet research. I also use Opencode a bit. All use cases I normally apply Opus to I've tried, as well as simpler prompts and mutli-step workflows. 35B seems to always perform as good or better and be much faster.

Edit:

35B is nvfp4 quant or sometimes fp8 and 27B is fp8 or nvfp4 quant

Edit 2:

I have 2 setups:

Home setup of Mac studio M4 Max 128Gb RAM, work mac M5 ~~ultra~~ max 48Gb ram.