Alibaba’s Qwen: The Chinese AI Model Challenging Silicon Valley

Researchers at Alibaba Group introduce Qwen, a family of open-source large language models trained on 3 trillion tokens. The 7B and 14B models power chat (via RLHF), coding (CODE-QWEN), math reasoning (MATH-QWEN), and vision-language (Qwen-VL). Benchmarks show performance rivaling GPT-3.5 in several tasks, strong tool use, and long-context handling—positioning Qwen as a serious global competitor in open AI.