Alibaba Group Holding Ltd. has released its largest-ever artificial intelligence model, claiming performance on par with global leader Anthropic PBC, marking the latest Chinese breakthrough challenging American competitors, UNN reports citing Bloomberg.
Details
The new Qwen3.8-Max is built on 2.4 trillion parameters and achieves higher scores on several tests than the recently unveiled well-known Kimi K3 from Moonshot. Alibaba shared results showing it delivers comparable or sometimes better results than Fable 5 from Anthropic, a frontier AI model whose export the US temporarily banned due to its advanced capabilities.
The debut comes days after Moonshot's Kimi sent ripples through stock markets and Silicon Valley, demonstrating that Chinese developers are rapidly catching up to the top models created by Anthropic and OpenAI, despite relatively limited computing resources. DeepSeek has also just expanded access to its latest V4 Flash model, while ByteDance Ltd. and MiniMax Group Inc. unveiled new video generators on Friday.
"Many investors continue to underestimate Chinese AI models due to US chip restrictions or general skepticism," said Wei-Sern Ling, managing director at Union Bancaire Privée.
"In reality, the gap is likely much smaller and narrowing quickly. Alibaba's Qwen 3.8 is another proof, after Kimi K3," he noted.
Alibaba shares rose as much as 7.3%, the biggest gain in nearly a month, on Monday in Hong Kong.
The Hangzhou-based internet pioneer will publish the weights of Qwen3.8-Max for public download next week, allowing users to customize the technology, marking the next major step in the intensifying race among Chinese AI competitors including Moonshot, DeepSeek, Z.ai, and ByteDance. While DeepSeek's latest release is the most affordable among new releases, Alibaba's Qwen offering also has an aggressive price — $2 per million input tokens and $6 per million output tokens. Each AI system will use a different number of tokens to perform tasks, but this still makes Alibaba's model attractive compared to the top models of US leaders, the publication writes.
The new Qwen3.8-Max performs well in autonomous programming and long-duration task execution, and according to Alibaba, was able to independently complete a software development project over 16 days during internal testing. The new system is also improved in efficiency, activating only some parts during operation to reduce computing costs and latency.
China considers strengthening export controls on AI models and chips - FT21.07.26, 12:32