Alibaba Drops 2.4-Trillion Parameter Qwen3.8-Max Weights and Enforces Revenue-Sharing License
In a major milestone for open-weight artificial intelligence, Alibaba Group Holding officially open-sourced the weights for its massive flagship model, Qwen3.8-Max (Qwen3.8-2.4T), on August 13, 2026. This marks the first "Max-scale" model in the Qwen series to have its weights fully released to the public. Alibaba is also scheduled to release Qwen3.8-27B on August 15, 2026, a smaller variant optimized for local deployment and edge devices.
Crucially, Alibaba is leveraging this release to pioneer a fundamental shift in the economics of open-weight models. Rather than treating open weights as free loss leaders, Alibaba is requiring large commercial users of Qwen3.8-Max to enter into revenue-sharing licensing agreements—following a precedent established by Moonshot AI's Kimi K3 license.
Technical Architecture of Qwen3.8-Max
Qwen3.8-Max is designed to compete directly with premier closed-source models like Google's Google DeepMind Launches Gemini 3.7 Flash at Half-Price as Flagship Gemini 3.5 Pro Delay Continues and OpenAI's GPT-5.5 series. Its technical specifications include:
- Total Parameters: 2.4 trillion.
- Active Parameters: 95 billion (utilizing a Mixture-of-Experts architecture).
- Context Window: 1 million tokens.
- Performance: Achieves an Intelligence Index composite score of 58, trailing only Anthropic's Claude Fable 5 (60) and matching or exceeding most other frontier models in coding and reasoning benchmarks.
The Revenue-Sharing Licensing Pivot
To extract commercial value from self-hosted deployments, Alibaba is implementing a tiered licensing structure that taxes large-scale enterprise usage:
- The Precedent: Moonshot AI's Kimi K3 License requires entities with a combined revenue exceeding $20 million in the trailing 12 months to share up to 30% of their revenue if they act as "Model as a Service" providers, or display attribution when exceeding 100M MAU or $20M in monthly revenue.
- Alibaba's Strategy: Large commercial users who self-host Qwen3.8-Max on their own infrastructure or GPU clouds must share a negotiated portion of their revenue with Alibaba. This represents a turning point in open-model strategy, establishing a "revenue ceiling" for free deployment and creating a sustainable commercial model for open-weights developers.