Testing 5 Qwen3.6-35B-A3B finetunes: the base model beats almost all of them

returnity · reddit · 2026-09-29

Reddit user returnity benchmarked the Qwen3.6-35B-A3B base model against five finetunes (Occamy-1.0, Ornith-1.5, KAT-Coder-V2.5-Dev, Tiel-Coder, Nex-N2.5-mini) on Aider Polyglot:

Chat template choice had minimal effect on scores.

Original post →

More from Models

Models channel →