Behavior of Mistral vs GPT-4?

Curious to hear user’s opinions of Mistral models vs OpenAI. Please add any anecdotal experience in this thread!

Hey - Two learnings:

  1. Visiting External Sites - ChatGPT is reluctant to visit websites. I know it can, but every so often it will say “I can’t do that” Mistral - happy to visit external websites although often overlooks details. Seems like it just gets high level idea and makes up the rest.

  2. Text Analysis - I’ve found Mistral to be superior on analysing transcripts compared to ChatGPT.

1 Like

Mistral does seem to have less “reinforcement learning” and ‘hard-coded personality’ than the OpenAI models. OpenAI often hedges against bad outputs by training their model to be more restrained.

Keep us updated! I myself am poking the models trying to find the key differences.