12

I think these models are outdated and should be either removed or moved to the bottom of the OpenAI list.

Someone who's not familiar with OpenAI models, would pick them by mistake. They cost more and have worse benchmark results than GPT 4o.

    slamor

    chatgpt-4-latest does not exist AFAIK, and the pricing you showed was for chatgpt-4o-latest (which I think is the model that they use on chatgpt.com)

    Correction:

    ModelInputOutput
    gpt-4$30.00 / 1M Tokens$60.00 / 1M Tokens
    gpt-4-turbo$10.00 / 1M Tokens$30.00 / 1M Tokens
    gpt-4o$2.50 / 1M Tokens$10.00 / 1M Tokens
    gpt-4o-mini$0.15 / 1M Tokens$0.60 / 1M Tokens

    To the suggestion: IMO these shouldn't be completely removed, but rather ranked like they did in the old assistant UI (with X/5 for speed X/5 for quality etc)

      16 days later
      5 days later

      I agree that these models shouldn't be removed as an option but some in-app indicator that explains that they're considered "legacy" could improve the experience of new users. Maybe a link to the benchmarking project?

        What's the reason for keeping the older models? GPT4o is much cheaper and better, and gets updated.

          They were removed entirely as of today and I want to oppose this removal. While they are slower and perform worse in general tasks, in very niche topics they often still give better answers because knowledge about these topics wasn't optimized/minified out.

          Yes. Agreeing with Lynatic here. I was pissed when I found out that GPT-4 was gone. I find GPT-4 to be significantly better in problem solving tasks.

          Agreed. It can be nice for certain tasks where it happens to provide more correct/predictable results. I'm all for warning the user about the choice they're making (worse performing on benchmarks, slower), but still leave the decision up to them! (you could even throttle legacy models further if it's too costly 😦, but leave them there! )

            No one is typing