Alibaba

Artificial Intelligence · Cloud Computing · Chinese Tech

Launch

Alibaba launches Qwen3.8-Max, a 2.4-trillion-parameter model rivaling Claude

Its open-weights release next week marks Alibaba's return to open-sourcing flagship AI after months of keeping top models closed.

• Alibaba made Qwen3.8-Max, its new flagship 2.4-trillion-parameter AI model, widely accessible via Alibaba Cloud's Model Studio APIs and its QwenWork platform, with an open-weights release planned for the following week.
• The model uses a sparse Mixture-of-Experts architecture that activates just 95 billion of its 2.4 trillion parameters, cutting compute costs while supporting a 1-million-token context window.
• Qwen3.8-Max ranked fifth on Arena.AI's Text Arena and second on its Vision Arena, trailing only Anthropic's Claude Fable 5 variants among global models.
• The launch follows Moonshot AI's 2.8-trillion-parameter Kimi K3 last month and DeepSeek's newly expanded V4-Flash, part of an intensifying race among Chinese labs to narrow the gap with US AI leaders.
• "Many investors continue to underestimate Chinese AI models because of US chip restrictions or general scepticism. In reality, the gap is probably much closer, and narrowing fast," said Vey-Sern Ling of Union Bancaire Privee.
• The release reinforces that Chinese developers can approach frontier AI performance despite export controls on advanced chips, a shift already reshaping how investors price Chinese tech stocks.