Comment by MangoCoffee
5 hours ago
cheap token is more important now than ever. Chinese open weight model is getting pretty good. the real cost of AI adaption will come down to who (China or US) can provide cheap token for consumers and companies. Microsoft consider DeepSeek for their cowork is an example and now OpenAI with its own AI inference chip.
I’m not understanding. If cost per token hits the floor that does not mean that you want a model that uses tokens.
If the Chinese are optimizing for token usage, that’s also speed.
Why use more token if few do trick?