This is really cool, nice work! Quick question - what would the cost of inference be, at scale, between a fine-tuned 3.5 and Llama 2 fine-tuned? Surely that's

This is really cool, nice work! Quick question - what would the cost of inference be, at scale, between a fine-tuned 3.5 and Llama 2 fine-tuned? Surely that's another factor that should be considered in this case, right?