Alibaba's Open-Source Qwen Models Surpass 3 Billion Downloads, Overtaking Meta in the On-Device AI Race

Alibaba's open-source AI model family Qwen has surpassed 3 billion downloads worldwide, overtaking Meta's Llama as the most-downloaded open-weight AI model. The Qwen 3.5 Small series (0.8B to 9B parameters) and the new Qwen3.8-27B are gaining traction for running AI locally on laptops and devices without GPU clusters.

Key Takeaways

Alibaba's open-source AI model family Qwen has surpassed 3 billion downloads worldwide, overtaking Meta's Llama as the most-downloaded open-weight AI model.
The Qwen 3.5 Small series (0.8B to 9B parameters) and the new Qwen3.8-27B are gaining traction for running AI locally on laptops and devices without GPU clusters.

AI synthesis3 sources6

Alibaba's open-source AI model family Qwen has surpassed 3 billion downloads worldwide over the past six months, overtaking Meta's Llama as the most-downloaded open-weight artificial intelligence model globally, according to recent reports.

The milestone comes as Alibaba has aggressively expanded its lineup of small, efficient models designed to run locally on consumer devices without requiring expensive GPU clusters. The Qwen 3.5 Small series — released in 0.8B, 2B, 4B, and 9B parameter variants — has proven particularly popular for on-device applications, offering lower inference costs and faster response times than larger cloud-based models.

Most recently, Alibaba launched Qwen3.8-27B, a laptop-ready open-weight model that the company says rivals Meta's best open models. The model has already generated over 151,000 derivative works from developers worldwide.

The surge in downloads reflects a broader shift in the AI industry toward smaller, locally-runnable models. As concerns about data privacy, latency, and cloud computing costs grow, developers and enterprises are increasingly looking for models that can operate on standard hardware rather than requiring specialized AI infrastructure.

Alibaba's Qwen models have found particular traction in China and Southeast Asia, where they power everything from chatbots to code assistants. The company has also been building partnerships with device manufacturers to embed Qwen models directly into smartphones and laptops.

The competition between Alibaba and Meta in the open-source AI space has intensified in 2026, with both companies releasing increasingly capable small models and vying for developer adoption. Meta's Llama family had long dominated the open-weight model downloads, but Qwen's rapid growth suggests the balance of power in open-source AI is shifting toward China.

The open-weight model market has become strategically important as governments worldwide debate AI regulation, with open models offering transparency and local deployment options that closed proprietary models cannot.