Our functions on GPT-4 work great, I tried with Llama-3-70b, but its doesn't seem to be calling yet, not sure if it's available at the moment since Llama-3 was just released a few days ago.