Ranking of top 500 GPTs by # conversations

https://gptranked.com/

This is the official data sourced from chat.openai.com - see how its done here: Custom GPTs — Only ~300 gets used out of 65K+ | by Gary Song | Dec, 2023 | Medium

GL all to the upcoming GPT store launch!

3 Likes

Very cool. Nice find. I’m confused why people are using a GPT model with a halved limit that’s… Just… Vanilla GPT? :thinking:

I’m a bit underwhelmed by the listing. Some cool ideas in there though

1 Like

How does this site scrape OpenAI and pull these stats?

1 Like

If there are, apparently, at the moment only a few hundred thousand chatgpt plus subscribers then I dont fully understand these numbers, unless they are used many times by a smaller population.

this blog explains how its done.

These numbers are by number of conversations, not unique users.

3 Likes

cool, and found a top 1000 gpts, https://www.gptsapp.io/trending-gpts/top-1000-gpts-ranked

The law of 2/8 is obvious, top is so high…

The latest GPT ranking GPT Store Top Ranking 500

1 Like

Agreed. If you look at the most popular GPTs (with the exception of the very top ones) they are mostly GPTs with just a catchy name (such as “Math”) and no added value (no knowledge files and no actions). They tend to have stellar reviews even though they are essentially a limited version of the basic chatGPT model. I think OpenAI should promote GPTs with added value more than those with no added value, but unfortunately it’s not the case for now.

A solution would be to allow users to search using keywords or GPT names. If you type “Math” in the GPT name box then you should see all GPTs with that name, but if you type it in the keyword box you should be displayed the GPTs that are actually more helpful to do math (regardless of their name).

2 Likes

What I dont understand is why making a sub copy of chatGPT? if people have to find ans they can use GPT itself why another copy or app which is doing same thing ?

Well first OpenAI has set the wrong incentives, whereby if your aim is to get a lot of users you should just call your GPT with a very popular name, without wasting time adding value. Then, I guess the average Chat GPT user does not understand GPTs as well as people in this forum, and thinks that a GPT called “Math” is a math super genius even though it is doing nothing different from the standard Chat GPT.

In my opinion if they don’t fix this soon they are going to lose the good developers to other platforms and retain only the low quality ones.

1 Like

Hey, I sent you a message since the ‘math’ GPT you’re referring to is mine and also thought it might be better to address this here too. I’ve noticed you made quite a few posts about it.

I’m curious what you think I can add to it to improve it since you seem to think that a much better GPT can be created. I get 10-20 feedback emails a day about it and most of them are overwhelmingly positive. For the ones that aren’t I iteratively update the prompt to fix their problem.

I’ve open sourced the prompt to my community of 3000 people and we’ve thoroughly refined the prompt and tested custom actions. Right now it seems pretty optimised and it’s not clear to me that using an api call for this GPT necessarily makes it better.

Any feedback is really appreciated, always looking for ways to make it more powerful.

Thanks,

Leo

1 Like

I used Math as an example, but I wasn’t referring to any GPT in particular. Indeed there are several GPTs called Math. My point is that if someone looks for “Math”, Wolfram should be the first GPT to come up, even though it’s name does not contain that string. But the same should apply to “Physics”, ecc…

I completely agree with you, and they went for a similar approach when the store was released by making your description determine what made you come up in search.

However people gamed that system and filled their description with a disgusting amount of key words so you’d have ones called ‘math physics science geography chemistry tutor assistant…’ all crammed into the description. You get the idea, it wasn’t great.

Adding an intelligent system which extracts the actual purpose of the GPT from the prompt itself rather than the name or description is a great approach in my opinion.

I don’t like that I have to call my GPT’s key words for names lmao but frankly its the only way anyone can actually find them with the current state of the store

I think tags chosen by the creator should put the GPT is a particular category; then what users see is determined by a function of rating and number of chats.