Alibaba's AI Model Qwen2-Math Surpasses Competitors in Mathematics

Alibaba introduces Qwen2-Math, an open-source AI model. Qwen2-Math achieves 84% accuracy in algebra and geometry problem-solving. Outperforms competitors like OpenAI's GPT 4o and Google's Gemini 1.5 Pro.

Credit: Alibaba

The tech giant claims that Qwen2-Math achieves an impressive 84% accuracy in solving algebra and geometry problems, surpassing the performance of OpenAI's GPT 4o and Google's Gemini 1.5 Pro.

This math-focused model, derived from Alibaba's foundation model Qwen2, was meticulously trained using a vast, high-quality math dataset. Currently available only in English, a bilingual version is in the pipeline, as revealed by Alibaba. The Alibaba team rigorously tested Qwen2-Math against various benchmarks, including challenging math problems from China's GaoKao (college entrance exam) and questions from the US math competition AIME. In all assessments, the model demonstrated exceptional proficiency in tackling complex mathematical challenges.

---

• Alibaba introduces Qwen2-Math, an open-source AI model

• Qwen2-Math achieves 84% accuracy in algebra and geometry problem-solving

• Outperforms competitors like OpenAI's GPT 4o and Google's Gemini 1.5 Pro

Source: TECHNODE