Beijing — In the intensifying technology race between China and the United States, Alibaba has made its boldest move yet. The Chinese e-commerce and cloud giant announced on Friday the release of its newest artificial intelligence model, Qwen3-Next-80B-A3B, claiming it to be not only faster than its competitors but also dramatically cheaper to produce.
The model, according to Alibaba Cloud, was developed at 90 percent lower cost than previous generations, yet still matches the power of far larger systems. At roughly one-thirteenth the size of Alibaba’s largest AI model launched just last week, the new system is positioned as a cost-efficient rival to American offerings.
India to Honour Top CISOs from Police, Law Enforcement, and Defence Forces
Built for Speed and Scale
A report in the South China Morning Post highlighted the breakthrough: the Qwen3-Next-80B-A3B runs ten times faster on some tasks than Qwen3-32B, a model introduced earlier this year. Industry observers note that Alibaba managed to bring development costs down to under ₹4.15 crore, compared to the ₹1,577 crore Google spent on its Gemini Ultra model.
“This is a clear demonstration of how China is looking to outmaneuver U.S. firms by focusing on efficiency, not just scale,” said Emad Mostaque, co-founder of Stability AI.
The Gated DeltaNet Advantage
At the heart of the model is a novel technology called Gated DeltaNet, first introduced by researchers at MIT and NVIDIA in March. The architecture enables the system to prioritize essential data while discarding irrelevant information, boosting speed and lowering resource demands.
Mixture-of-Experts Design
Alibaba also incorporated a Mixture-of-Experts (MoE) framework, splitting the model into 512 specialized units, only 10 of which are active at any given time. This design mirrors strategies adopted in other Chinese models like DeepSeek-V3 and Kimi-K2, but on a larger scale.
A Cheaper Challenger to American Giants
For comparison, Elon Musk’s xAI reportedly spent more than ₹4,070 crore to develop its Grok-4 system. By contrast, Alibaba’s new model can run on a single NVIDIA H200 GPU, a chip that costs roughly ₹24.9 lakh. That makes it both more accessible and easier to deploy. Within 24 hours of its launch, the model had been downloaded more than 20,000 times on Hugging Face, the open-source AI platform.
A Symbol of China’s AI Strategy
The release underscores Beijing’s determination to challenge U.S. dominance in artificial intelligence. Instead of building only the biggest models, Chinese companies are focusing on leaner, cheaper, and highly efficient systems — a shift that could reshape the economics of AI development worldwide.
“Alibaba’s announcement is not just about performance benchmarks,” said a Beijing-based AI policy analyst. “It is a statement that China intends to compete in AI on its own terms — speed, cost, and accessibility — rather than playing catch-up with America’s trillion-parameter models.”