The extension with gpt4 as a backend was ime extremely slow as standard. I've not tried it again with the v7 model though which is supposed to be a lot faster