Comment by ein0p

This is from a small model. 32B and 70B answer this correctly. "Arrowroot" too. Interestingly, 32B's "thinking" is a lot shorter and it seems to be more "sure". Could be because it's based on Qwen rather than LLaMA.