At the World AI Conference in Shanghai, Alibaba announced the preview release of its Qwen3.8-Max AI model, asserting it as one of the most powerful models worldwide, second only to Anthropic’s Claude Fable 5. This milestone highlights the rapid advancement of Chinese AI technology into the multi-trillion parameter era.
- Qwen3.8-Max-Preview has 2.4 trillion parameters, ranking just below Claude Fable 5.
- Moonshot AI’s Kimi K3 model leads the open-source field with 2.8 trillion parameters.
- China's efforts reflect a strategic push to rival US AI leaders like OpenAI and Anthropic.
What happened
Alibaba Group announced its next-generation AI model, Qwen3.8-Max-Preview, available via its Token Plan subscription and AI platforms Qoder and QoderWork. The model boasts 2.4 trillion parameters, placing it among the world’s most powerful AI systems. Alibaba claims it is second only to Anthropic’s Claude Fable 5, marking a significant advance in Chinese AI capabilities.
The revelation came during the World AI Conference in Shanghai, underscoring China's arrival in the multi-trillion parameter AI realm. Alongside Alibaba, Beijing-based Moonshot AI debuted their Kimi K3 model with 2.8 trillion parameters, currently the largest open-source AI model globally. This model outperforms several leading proprietary systems in various benchmarks, indicating vigorous innovation competition within China.
Why it matters
This development signals China's accelerating drive to compete with US AI powerhouses such as OpenAI and Anthropic by developing models with significantly increased complexity and capabilities. High parameter counts generally correlate with more sophisticated AI performance, and these advancements position Chinese entities to challenge Western dominance in AI development.
Beyond technical achievement, these efforts are aligned with national strategic priorities. President Xi Jinping emphasized the importance of openness, collaboration, and open source approaches to foster innovation and maximize AI’s industrial and societal benefits. By advancing frontier AI models, Chinese companies and government aim to strengthen domestic tech leadership and global influence.
What to watch next
Alibaba plans to release Qwen3.8 as an open-weight model, sharing the model's intelligence embeddings while withholding full training data or source code. This may expand accessibility for developers and accelerate ecosystem growth around Qwen3.8. Independent benchmarking and comparisons to Moonshot’s K3 and other competitors will be crucial to validate Alibaba’s performance claims.
Additionally, the competitive landscape in China is intensifying with several labs introducing increasingly large and capable AI models. The global AI community will observe how these Chinese prototypes compete against US firms’ upcoming models in performance, application breadth, and commercialization, shaping the future dynamics of AI leadership and innovation.