Alibaba Group BABA released its latest artificial intelligence model on Friday. It unveiled Qwen3-Next, an architecture designed to deliver ten times more performance at just a tenth of the cost of its predecessor.
The Qwen team explained on GitHub that it integrated multiple architectural innovations to maximize efficiency while reducing computing needs.
Developed by Alibaba Cloud, the model, Qwen3-Next-80B-A3B with 80 billion parameters, costs far less to train and runs up to ten times faster than the earlier Qwen3-32B launched in April, SCMP reported on Friday.
Also Read: Chinese Regulators Question Nvidia Orders But Alibaba And ByteDance Keep Buying
Alibaba Cloud noted that Qwen3-Next matched the performance of its flagship Qwen3-235B-A22B while being optimized for deployment on consumer-grade hardware.
The team also released a reasoning-focused version, Qwen3-Next-80B-A3B-Thinking, which outperformed Qwen3-32B-Thinking and Alphabet GOOGL GOOG Google’s Gemini-2.5-Flash-Thinking in third-party benchmarks.
The launch underscores intensifying competition in China’s AI sector. DeepSeek is developing agent-based models for autonomous multi-step tasks, while Meituan MPNGY is advancing open-source research with its 560B-parameter LongCat-Flash-Chat.
These divergent strategies reflect China’s accelerating race to challenge OpenAI, Microsoft MSFT, Alibaba, and Tencent TCEHY in shaping the next wave of AI innovation.
Alibaba shares have surged more than 83% year-to-date, outpacing the NYSE Composite Index’s 13% return as its AI initiatives gain traction. The company has built Qwen into the world’s largest open-source AI developer ecosystem, helping to narrow the gap with U.S. rivals.
Price Action: BABA stock is trading lower by 0.39% to $154.83 premarket at last check Friday.
Read Next:
Photo by Poetra.RH via Shutterstock
© 2025 Benzinga.com. Benzinga does not provide investment advice. All rights reserved.
