GPT-4 analyzing online image

Hey fam,

First of all, I am not sure whether I have access to the correct model.

After I received the email suggesting the waiting is over, I now can see the GPT-4 in Playground.

However, in the API call, if I change the model to “GPT-4”, it says “Model cannot be found”. But If I change to “gpt-4-0314” (based on some posts suggesting this trick), I could use the model.

But when I provide the user message like “Can you analyse this image URL”, it suggests cannot access the image as a language model.

Then if I change the model back to “GPT-4”, the API respond back as “gpt-4-0314” still.

Am I doing something wrong here?


apparently it’s not yet possible

1 Like

Hello @nhtkid and @RePattern
I came across your same problem. I have an App working with gpt3.5 and I switched to gpt4, asking the model to describe an image for which I provided an http link. The model replied with a message suggesting it cannot access the image as a language model and it can process only text. But if I try the same link on my OpenAI Playground, it works and gpt4 is able to describe the image. Probably this feature is not yet available for developers working with the GPT4 API?

1 Like