Today, researchers from Stanford University and Washington University, including Li Feifei, trained an artificial intelligence inference model called s1 for less than 50 US dollars in cloud computing costs. The model performed similarly to cutting-edge inference models such as OpenAL's O1 and Depsek's R1, which attracted widespread attention. Soon, however, the S1 model was accused of “not being trained from scratch,” and its base model was the “Ali Tongyi Thousand Questions Model.” In response, Sina Technology sought evidence from Alibaba Cloud. Alibaba Cloud confirmed the news and responded, “Using the Alibaba Tongyi Qianwen 2.5-32B-Instruct open source model as a base, they supervised and fine-tuned for 26 minutes on 16 H100 GPUs, and trained the new model S1-32b, which achieved results comparable to the mathematical and coding abilities of cutting-edge reasoning models such as OpenAI's O1 and DeepSeek's R1.”

Today, researchers from Stanford University and Washington University, including Li Feifei, trained an artificial intelligence inference model called s1 for less than 50 US dollars in cloud computing costs. The model performed similarly to cutting-edge inference models such as OpenAL's O1 and Depsek's R1, which attracted widespread attention. Soon, however, the S1 model was accused of “not being trained from scratch,” and its base model was the “Ali Tongyi Thousand Questions Model.” In response, Sina Technology sought evidence from Alibaba Cloud. Alibaba Cloud confirmed the news and responded, “Using the Alibaba Tongyi Qianwen 2.5-32B-Instruct open source model as a base, they supervised and fine-tuned for 26 minutes on 16 H100 GPUs, and trained the new model S1-32b, which achieved results comparable to the mathematical and coding abilities of cutting-edge reasoning models such as OpenAI's O1 and DeepSeek's R1.”