Alibaba Group Holding Limited

New York Stock Exchange
Back to all articles
Slightly Bullish +20

Alibaba previews Qwen3.8, claims it's second only to Claude Fable 5

🚀 Alibaba previews Qwen3.8-Max-Preview with 2.4 trillion parameters, capable of processing text, images, video, and documents.

📊 Company claims the model ranks second only to Anthropic's Claude Fable 5 among world frontier models.

⚠️ No benchmark scores, model cards, or independent evaluations were provided for the Qwen3.8 preview.

💰 Preview available via Token Plan subscription at 10% of standard rate during trial period.

🇨🇳 Launch occurs three days after Beijing Moonshot AI released its 2.8 trillion-parameter Kimi K3 model.

🔓 Open weights promised for Qwen3.8, though no specific date or license terms have been set.

🏢 Alibaba positions Qwen as central to its strategy of becoming China's default AI supplier.

📉 Previous flagship Qwen3.7-Max scored 56.6 on Artificial Analysis Intelligence Index with full results published.

🤝 Model available through Qoder and QoderWork developer platforms for enterprise buyers.

Bullish Signals
  • Alibaba has successfully developed a model with 2.4 trillion parameters, entering the multi-trillion-parameter tier previously associated mainly with large U.S. labs.
  • The company maintains a strong developer following built on its history of releasing open-weight models, which it hopes to carry into the trillion-parameter class.
  • Qwen3.8 is offered at a discounted rate of 10% during the trial period, making advanced AI capabilities more accessible for developers and enterprises.
  • Alibaba continues to build out cloud infrastructure and custom chips to support its high-capacity AI models, reinforcing its technical ecosystem.
Risk Factors
  • The company provided no benchmark scores or independent evaluation data to support its claim that Qwen3.8 ranks second only to Claude Fable 5.
  • Alibaba published no task-level comparison with its previous flagship Qwen3.7-Max, describing the new model only as 'continuously evolving' without specific metrics.
  • The absence of a model card and activated-parameter count for Qwen3.8 contrasts sharply with the full set of published results provided for Qwen3.7-Max in May.
  • Enterprise buyers face uncertainty regarding the model's actual performance relative to competitors like Fable 5 due to the lack of published benchmarks or leaderboard rankings.
  • Open weights are promised 'soon' but no date or license terms have been published, creating potential delays for developers seeking open-source access.
Full Analysis
Alibaba Group Holding Ltd. previewed Qwen3.8-Max-Preview, a large language model with 2.4 trillion parameters capable of processing text, images, video, and documents. The company claims the model ranks second only to Anthropic's Claude Fable 5 among global frontier models, positioning it as the most powerful in its family. Unlike previous releases that included benchmark scores and model cards, Alibaba provided no independent evaluation data or specific performance metrics for Qwen3.8. The preview is available via subscription at a discounted rate of 10% during the trial period, with open weights promised for an unspecified future date. The launch coincides with Beijing Moonshot AI Technology Co. Ltd.'s release of its 2.8 trillion-parameter Kimi K3 model three days prior, signaling that Chinese developers have entered the multi-trillion-parameter tier previously dominated by U.S. labs. Alibaba aims to solidify its position as China's default AI supplier through this high-capacity model. Enterprise buyers face uncertainty regarding the model's actual performance relative to competitors like Fable 5 due to the lack of published benchmarks or leaderboard rankings. While Alibaba has built a strong developer following with open-weight models, the absence of verification data for this specific preview limits immediate comparative assessment.