Alibaba shares jump on unverified claim of second-best AI model
Alibaba has previewed a massive new AI model it claims trails only Anthropic’s best, but the absence of public testing data leaves the market waiting to see if China's open-source push is real.
Alibaba has unveiled Qwen3.8 at the World Artificial Intelligence Conference in Shanghai. The Chinese tech giant claims the new system is “second only to Fable 5,” the top-ranked model from Anthropic. The bold assertion pushed Alibaba shares up as much as 5.4% on Monday.
The model holds 2.4 trillion parameters, marking the first time a Qwen system exceeding a trillion parameters can process images, video, and documents alongside text. Lead developer Shuai Bai said the model's understanding matches or beats leading proprietary systems. It is designed to outperform its predecessor at full-stack development and data analysis.
However, the company has provided no benchmark scores, model card, or independent tests to justify the ranking. This departs from the release of its previous flagship, Qwen3.7-Max, which shipped in May with full public results. On the LMArena leaderboard where Fable 5 sits at number one, the older Qwen model ranks well down the list.
The timing is a direct response to escalating competition. The launch arrived three days after Chinese rival Moonshot released Kimi K3, a 2.8-trillion-parameter model that topped a major coding leaderboard and rattled US chip stocks. A cluster of Chinese labs, including Zhipu with its GLM-5.2, are now releasing models at a scale previously dominated by US firms.
For investors and enterprise users, the strategic rollout matters as much as the technology. Alibaba is offering the preview through its Token Plan subscription and Qoder developer tools at 10% of the standard rate. By monetizing the system now and promising open weights later, the company captures customers before anyone can run the model for free.
The broader significance lies in Alibaba abandoning its usual reluctance to open-source its largest models. AI researcher Nathan Lambert observed that “Something is changing,” identifying a new phase of Chinese lab competition. If the eventual weight release validates the hype, it would push open-source AI deep into frontier territory, challenging a defensive moat that US labs have fiercely protected.
