Is deepseek’s r1 model distillated from ChatGPT?

Yes, there is evidence suggesting that DeepSeek’s R1 model was developed using a technique called “distillation� from OpenAI’s models. Distillation involves training a new model by leveraging the outputs of a pre-existing model, effectively transferring knowledge from the original to the new model. OpenAI has indicated that it found evidence linking DeepSeek to the use of distillation—a common technique developers use to train models by leveraging the outputs of existing ones. 

 

Additionally, discussions within the AI community have raised concerns about DeepSeek’s methods. For instance, a thread on the OpenAI Developer Forum titled “Looks like Deep Seek R1/V3 was distilled from GPT-4/3.5 - Can anyone confirm?� delves into this topic. 

 

Therefore, it appears that DeepSeek’s R1 model was indeed developed through distillation from OpenAI’s models.