Institution: Distillation technology significantly reduces development costs and is expected to accelerate the practical implementation of AI applications

The Guosheng Computer team stated that DeepSeek distilled six small models from the output of DeepSeek-R1 and opened them up to the community. Among them, the 32B and 70B models achieved benchmarking effects against OpenAI o1-mini in multiple capabilities, and the DeepSeek-R1 API service is priced much lower than OpenAI o1. Lower costs enable enterprises to develop AI applications with higher efficiency, which is expected to accelerate the transition of domestic AI applications from concept to practical implementation. DeepSeek’s open-source distillation model surpassing OpenAI o1 mini is also expected to accelerate the model’s implementation on the end side.