Day of the week
Thank you for this...verbose...model. The first question I typically ask a new model is which day of the week it is, given a calendar day, and this model correctly deduced it the first time. I've found that prior (and smaller) models struggle with this specific task. I used a temperature setting of 0.2.
However, it used significantly more electricity than phi-4 and Llama-3.3-70B-Instruct to reach the answer: these models reach the answer quickly. The reasoning step was chasing its own tail for a while: it was circling between Zeller's and the Doomsday method, because it continued to second-guess if it had either formula or method correct. During the reasoning step, it correctly emitted Zeller's method and the Doomsday method the first time, so the repetitive second-guessing was wasteful.
I've saved the output, if anyone is interested I can post it, but it is very long.
I'll try to employ this for other tasks, thanks again!