#when will custom GPTs will have reasoning o3 mini model?

1 messages · Page 1 of 1 (latest)

clever brook
brazen kindle
#

I have the same doubt the same doubt you have.

modern yew
#

projects folders solve that issue and its even better than custom gpts

ocean schooner
# modern yew projects folders solve that issue and its even better than custom gpts

I wouldn't call Projects better; they are a bit different, and in fact, basically a fork of Custom GPTs...

Custom GPTs are more like characters with stuff like Name, Icon, and Description. Multiple different GPTs can be mentioned in the same chat with the @ command (which I found to be quite useful). As well as Action calling (API interactions). All of which makes them more fit for the assistant / agent / character with personality role.

modern yew
#

Yeah but in practice projects are much better because they have way better context window and more custom instructions

clever brook
modern yew
ocean schooner
#

Custom GPTs always had RAG with the files uploaded (the knowledgebase). In terms of projects, that's the same thing, but it doesn't use context from other chats in a project. If it does for you, it may be the new advanced memory feature which can retrieve information from past convos, which is not related to Projects and more of ChatGPT as a whole (and btw, if you have it enabled, I would recommend disabling it to avoid context pollution)

clever brook
ocean schooner
clever brook
main python
#

Custom GPTs require a paid subscription to create, the unique thing they offer is Actions, i.e. you can make them Agents if you also build out the API. You can also publish them publicly. Projects are personal only, accessible with a free account and allow you to group similar conversations. It would be nice if you could add custom GPTs to your projects, but alas, you cannot. Custom GPTs are by far superior. I'm sure there are many projects where Projects are sufficient and loads where there is overlap, and if Projects work better for you in that regard, that's cool.