Data, model arch, compute, and post training processing. I’m assuming all bigModel cos have good data and compute access which means the model arch and post pr

Data, model arch, compute, and post training processing. I’m assuming all bigModel cos have good data and compute access which means the model arch and post processing is where the differentiation is at? I know OpenAI is augmenting with function calling techniques. Where is the real differentiation ? Why is open ai so much better?