Alibaba unveiled its latest and most powerful AI model, Qwen3.8-Max, which it claims rivals the performance of top systems from US companies like Anthropic and OpenAI, as well as domestic competitors such as Moonshot AI's Kimi K3. The model features 2.4 trillion parameters, a numerical measure of its learning capacity, and supports a context window of up to 1 million tokens, allowing it to process extensive information equivalent to thousands of pages. This release marks Alibaba's return to open-weight models, providing developers with more control compared to proprietary systems.
Alibaba's internal testing indicates Qwen3.8-Max performs comparably to, and sometimes exceeds, Anthropic's Fable 5 on benchmark tests. On crowdsourced model-comparison platforms, it ranks as the highest Chinese model for text and second globally for visual analysis, trailing only certain Claude Fable 5 and Opus variants. The model is designed with a "mixture-of-experts" architecture, using only 95 billion parameters at a time to reduce costs and latency, and can handle text, images, and video.
The launch of Qwen3.8-Max had a significant impact on Alibaba's stock, with its New York-listed shares climbing 4.5% and Hong Kong shares rising 7% on Monday. The company priced the model competitively at $2 per 1 million input tokens and $6 per 1 million output tokens. Alibaba also introduced QwenWork, a workplace AI platform in public beta, which integrates existing tools and aims to bring Qwen models into enterprise productivity workflows. Analysts from Susquehanna, JPMorgan, and Barclays maintained positive ratings, with price targets ranging from $185 to $205, citing the launch as strengthening perceptions of Alibaba's AI competitiveness.
