Alibaba has announced the release of Qwen3, a new family of AI models designed to rival top models from Google and OpenAI. The models, ranging from 0.6 billion to 235 billion parameters, will be available under an open license on Hugging Face and GitHub.
Alibaba described Qwen3 as “hybrid” models capable of balancing fast responses with deeper reasoning when needed. The company said Qwen3, trained on over 36 trillion tokens, supports 119 languages and significantly outperforms its predecessor, Qwen2.
CEO Tuhin Srivastava of Baseten noted that models like Qwen3 highlight how open-source AI is increasingly competitive with closed systems. Some Qwen3 models are already accessible via cloud providers, further accelerating global AI adoption.
