Alibaba’s Qwen3 Overtakes DeepSeek’s R1 as Top Open-Source AI Model

Alibaba Cloud’s newly launched Qwen3 family of AI models has dethroned DeepSeek’s R1 as the world’s leading open-source artificial intelligence framework. Benchmark results from independent evaluator LiveBench show superior performance across language instruction, mathematical reasoning, coding tasks, and data analysis capabilities.

The eight-model Qwen3 series, ranging from 600 million to 235 billion parameters, represents Alibaba’s latest advancement in machine learning infrastructure. Parameter count—the adjustable variables during AI training—has become a critical metric for model sophistication in the competitive AI development landscape.

This technological shift comes just months after DeepSeek’s R1 claimed the open-source crown in January. The rapid succession of breakthroughs underscores the intensifying AI arms race between Chinese and American tech firms, with both nations’ enterprises reportedly rushing to implement Qwen3 architectures.