ParentFull threadbigyabai·Presumably because GLM 4.5 or Qwen3 comparisons would clobber them in eval scores.View on HN