Recently I have been trying to make an app using the GPT-4 model - to ensure I got it correctly working, I asked what model this is using. It responded with GPT-3. When I asked the same question in the openai chatgpt portal - it did know it was using the gp4 model and responded as such.
The GPT models are trained on data that ends prior to the models creation, the only way the models know what version they are is by system prompts that tell them this fact. GPT AI’s are not self learning or auto updating, they use a fixed dataset that occasionally gets updates, any attempt to ask a model what version it is will typically be unreliable.
When using the API, you use system message programming to tell the AI what it is and how you’d want it to respond, including the name of the AI model it reports.
