Alibaba Group's (HKG:9988) shares surged 8% after it introduced QwQ-32B, an open-source AI reasoning model it claims performs better than DeepSeek's R1 on Thursday, the SCMP reported the same day.
The QwQ-32B model outperformed R1 in areas including problem solving, mathematics, and coding, SCMP said citing a blog post by the team responsible for the Qwen AI model family by Alibaba.
Compared with R1's 671 billion parameters, Alibaba's new model boasts 32 billion, which the same team claims allows it to use fewer resources for computation.
Alibaba also said its new model outshined OpenAI's o1-mini, with the team attributing its boost in performance mainly to reinforcement learning techniques, according to the report.
(Market Chatter news is derived from conversations with market professionals globally. This information is believed to be from reliable sources but may include rumor and speculation. Accuracy is not guaranteed.)