Alibaba (Qwen)

Alibaba's AI division and the Qwen (通義千問) open-source model family — currently the most influential Chinese open-source LLM platform, with rapid release cadence.

Alibaba's AI work runs through Alibaba DAMO Academy and Tongyi Lab, with the consumer-facing model line branded Qwen (通義千問, "a thousand questions of universal meaning"). The team has shipped models continuously since 2023 — Qwen 1, 1.5, 2, 2.5, 3, with sizes from 0.5B parameters up to 235B+ MoE — and almost all weights are released openly under permissive licenses. It matters because Qwen is currently the most-downloaded Chinese open-source LLM family on Hugging Face. For Chinese-language work, it's often the strongest open-weight option. The release cadence is faster than any Western open-source effort: new sizes, multimodal variants, code variants, agent variants ship every few months. Key contributions: the full Qwen line (general chat), Qwen-VL (vision), Qwen-Audio, QwQ (reasoning), Qwen-Coder, and a 1M-context variant. Alibaba also runs Tongyi Qianwen as a hosted product and powers Bailian (its enterprise AI platform) and ModelScope (China's Hugging Face equivalent). For Chinese-market builders, Qwen is the default "open-source frontier" — comparable to Llama for English work. Western adoption is also growing as Qwen frequently competes with Llama on global benchmarks. Related: Qwen family, open-source, ModelScope, Llama family.