Does DeepSeek own enough compute power to actually leverage the higher efficiency of this model? Doesn’t help if it’s cheaper on paper in small scale, if you physically don’t have the capacity to sell it as a service on a large scale.
By the time they do have the scale, don’t you think OpenAI will have a new generation of models that are just as efficient? Being the best model is no moat for any company. It wasn’t for OpenAi (and they know that very well), and it’s not for Deepseek either. So how will Deepseek stay relevant when another model inevitably surpasses them?
By the time they do have the scale, don’t you think OpenAI will have a new generation of models that are just as efficient? Being the best model is no moat for any company. It wasn’t for OpenAi (and they know that very well), and it’s not for Deepseek either. So how will Deepseek stay relevant when another model inevitably surpasses them?