Simplifying Model Version Management with GPT-latest

GPT-4-Turbo always points to the latest and best version of GPT-4-Turbo, which is convenient. However, when GPT-4o is released, you have to change everything to GPT-4o. This will keep happening with each new model.

To solve this, we need a general model pointer, like ‘GPT-latest,’ that always leads to the newest and best model available. This way, those who prefer manual control over model upgrades can continue using the specific model pointers, while others benefit from automatic updates.

You would not want API applications to point to gpt-4o automatically. It is a significantly different model, and being cheaper shows in its capabilities.

1 Like

This was not true in our case and we had to manually update the model name in doznes of places. (We could solve this with a workaroung but the need is there).

In the case you are describing, one can simply continue using the model-specific pointers.

1 Like

A pointer within your application would be sufficient though, wouldn’t it?
What stops you from declaring your own constant GPT_LATEST, use it everywhere you need and only update that constant whenever you want?

1 Like

This looks needed and we might do so. However, our system is built across several different platforms, there isn’t a single codebase, and in some parts of it this approach won’t work as constants are unavailable.

Having a solution straight from OpenAI will probably save humanity a couple of human hours dealing with this issue.