Alibaba Qwen Hits 3 Billion Downloads, Overtakes Meta and Google: Inside the August 2026 Open-Weight AI Landslide

The Open-Weight Releases: Qwen3.8-Max and Qwen3.8-27B

The 3-billion-download milestone was supported by two major open-weight releases in August 2026: the flagship Qwen3.8-Max and the smaller Qwen3.8-27B.

Qwen3.8-Max (Qwen3.8-2.4T-A95B) — first launched as an API product on August 3, 2026, with its weights published openly on Hugging Face on August 13, 2026 under a custom Qwen license . This was the first time Alibaba released weights for a Max-class model . Key specifications include:

• 2.4 trillion total parameters, sparse Mixture-of-Experts (MoE) architecture with roughly 95 billion active parameters per token
• 1-million-token context window (991K max input, 131K max output)
• Native multimodal input (text + vision) with hybrid thinking enabled by default

Qwen3.8-27B — the smaller dense companion model, with weights released on August 13-14, 2026 on Hugging Face under the Apache 2.0 license . This model is designed for self-hosting on a single GPU and includes a native vision encoder with a 262K token context . Alibaba reported that Qwen3.8-27B scored 61.7 on SWE-bench Pro and 84.3 on OSWorld-Verified, outperforming larger proprietary models on several agentic coding benchmarks .

Competitive Context: Overtaking U.S. Giants

Qwen's dominance signals a notable shift in developer preference:

• Vs. Meta: Qwen overtook Meta's Llama family as the most-downloaded open-weight ecosystem, with 3 billion Qwen downloads versus 227 million for Meta .
• Vs. Google: Alibaba surpassed Google's Gemma/Gemini open-weight offerings, which recorded 418 million downloads .
• Vs. Anthropic: Alibaba positioned Qwen3.8-Max as competitive with Anthropic's Fable-class models in performance benchmarks .
• Vs. Moonshot AI: Domestic rival Moonshot had launched the Kimi K3 model (2.8 trillion parameters) just before; Alibaba's Qwen3.8-Max was described as being "not far behind" Moonshot's model in size, but ranking higher on several benchmarks .
• Stock Impact: Alibaba shares jumped 7% in Hong Kong trade on the August 3 Qwen3.8-Max unveiling .

Revenue-Sharing Plans for Large Commercial Users

On August 7, 2026, Reuters reported that Alibaba planned to require large commercial users of Qwen3.8-Max open weights to share a portion of their revenue generated from products or services built using the model . Key details:

• The plan targets major enterprise users, not researchers or small-scale developers .
• The exact revenue-sharing percentage had not been finalized as of the reporting, with negotiations still ongoing .
• Alibaba followed a precedent set by Chinese rival Moonshot AI, whose Kimi K3 license reportedly sought a revenue share of up to 30% for companies with annual revenue above roughly $20 million .
• The terms were expected to take effect around the open-weight release in the week of August 10 .
• This move was described as a pioneering shift in open-weight AI economics — a monetization layer on top of ostensibly open weights, fundamentally altering the "free open-source AI" model for large-scale commercial use . Alibaba had previously charged for model use on its own cloud but generally allowed most open-weight releases to run in customer data centers without additional fees .