Custom "high" reasoning mode for 27B model blends low and xhigh prompts

TokenRingAI · reddit · 2026-08-15

After discovering that a 27B model's "medium" reasoning mode ignores instructions and degrades performance, the author experimented with blending keywords from "low" and "xhigh" prompts. The resulting "high" mode generates reasoning blocks approximately 1/5 the size of xhigh while maintaining similar output quality. The post provides the specific chat template modification to implement this custom reasoning effort.

Original post →

More from Models

Models channel →