Alibaba Launches New Qwen LLMs

Source Credit : CNBC

Alibaba unveiled the latest iteration of its open-source large language models, Qwen3, on Tuesday, marking a significant advancement in China's rapidly growing open-source artificial intelligence sector. In a recent blog post, the Chinese tech powerhouse highlighted Qwen3's enhancements in reasoning, instruction comprehension, tool utilization, and multilingual capabilities, positioning it as a formidable competitor to other leading models like DeepSeek's R1 in various industry benchmarks.

The LLM series comprises eight variations that encompass a variety of architectures and sizes, providing developers with flexibility when utilizing Qwen to create AI applications for edge devices such as mobile phones.

Furthermore, Qwen3 marks Alibaba's foray into hybrid reasoning models, which combine traditional LLM capabilities with advanced, dynamic reasoning.

Alibaba has stated that these models have the ability to smoothly switch between a "thinking mode" for intricate tasks like coding and a "non-thinking mode" for quicker, general-purpose responses.

Alibaba emphasized that the Qwen3-235B-A22B MoE model notably reduces deployment costs in comparison to other cutting-edge models, demonstrating Alibaba's dedication to providing accessible, high-performance AI solutions.

The latest models are now readily accessible to individual users on platforms such as Hugging Face and GitHub, as well as through Alibaba Cloud's web interface. Additionally, Qwen3 is currently being utilized to enhance Alibaba's AI assistant, Quark.