I would think there’s a major difference in the inference time commute for tools like this. And major providers can spend a lot more (at a loss) on the runtime

I would think there’s a major difference in the inference time commute for tools like this. And major providers can spend a lot more (at a loss) on the runtime compute. That’s just a guess though.