Alibaba has unveiled its QwQ-32B-Preview artificial intelligence (AI) model, which reportedly outperforms OpenAI's o1 series. The model surpasses OpenAI on specific benchmark tests, such as logic puzzles and math problems. Alibaba's model can handle complex problems and has 32.5 billion parameters, allowing it to handle prompts of up to 32,000 words. The model is available for users to download and use, but certain components have been withheld. While the model has strengths in reasoning, it has limitations such as unexpected language switches and underperformance in common-sense reasoning tasks. However, it can fact-check itself and aligns with Chinese regulation on politically sensitive topics. The model's launch positions Alibaba as a competitor to OpenAI in the AI reasoning technology field.



Other News from Today