It's fine not to give them public credit for in-house only things, but in this subthread we're speculating about whether Llama 2 would be useful to them, which does depend heavily on the quality of their internal models.