Is davicinci-instruct-beta-v3 cheaper? or is it just prompt efficiency?

I’m working on building synthetic datasets for an experiment and I swear that this model is consuming less money. I know some of the cost difference is the fact that I’m using instruct rather than vanilla with few-shot learning. My instruct prompts are pretty lean.

If instruct and vanilla are the same price then holy smokes, instruct is like 5x cheaper with lean prompts!

I feel like instruct-beta-v3 is much better at following simple instructions, hence why the prompts are simpler. I’d be curious if there’s a publication somewhere listing the improvements of v3.

4 Likes