In short, LLaMA's focus is on 'how much performance you can squeeze out for the cost', while OpenAI is more about 'how low they can get the price for high performance'. OpenAI values scalability a lot, so GPT models are easy to scale up and use across all kinds of fields. On the other hand, LLaMA feels more like something fun to tinker with when you want to accomplish a specific task.