The Blubber

Alibaba Unveils Qwen 2.5 AI Model, Challenging DeepSeek’s Rise

Chinese tech giant Alibaba (9988.HK) released an updated version of its Qwen 2.5 artificial intelligence model on Wednesday, claiming it outperforms the widely praised DeepSeek-V3.

The timing of the Qwen 2.5-Max launch—on the first day of Lunar New Year, when most people in China are off work—highlights the intense competition DeepSeek’s meteoric rise has sparked, not just against overseas AI developers but also within China’s domestic AI industry.

“Qwen 2.5-Max outperforms GPT-4o, DeepSeek-V3, and Llama-3.1-405B across almost all benchmarks,” Alibaba’s cloud division announced on its official WeChat account, referring to OpenAI and Meta’s latest AI models.

DeepSeek's Disruptive Impact

The January 10 release of DeepSeek’s AI assistant, powered by DeepSeek-V3, and the subsequent January 20 launch of its R1 model, sent shockwaves through Silicon Valley. The startup’s low-cost AI model led to a global tech sell-off, with investors questioning the massive spending plans of leading U.S. AI firms.

But DeepSeek’s rapid ascent has also fueled an AI arms race in China, with competitors scrambling to enhance their own models.

Just two days after DeepSeek-R1 debuted, ByteDance, the owner of TikTok, released an updated flagship AI model, claiming it outperformed OpenAI’s o1 model in AIME, a benchmark test measuring AI’s ability to understand and execute complex instructions.

DeepSeek had already made similar claims about its R1 model, suggesting it could rival OpenAI’s o1 in several key performance metrics.

DeepSeek’s Challenge to Chinese Tech Giants

The impact of DeepSeek’s AI models goes beyond performance—it has triggered a pricing war in China.

When DeepSeek-V2 was released last May, its open-source framework and ultra-low pricing—just 1 yuan ($0.14) per million tokens—forced Alibaba Cloud to slash prices on its own AI services by up to 97%.

Other Chinese tech giants quickly followed suit:

• Baidu (9888.HK), which introduced China’s first ChatGPT-like model in March 2023, responded with its own price cuts.
• Tencent (0700.HK), China’s most valuable internet company, also adjusted its AI pricing strategy.

Despite the market shakeup, DeepSeek's elusive founder, Liang Wenfeng, has remained focused on his ultimate goal: achieving Artificial General Intelligence (AGI).

In a rare July interview with Chinese media outlet Waves, Liang dismissed pricing wars, stating, "We don’t care about price competition. Our goal is AGI."

OpenAI defines AGI as AI systems capable of outperforming humans in most economically valuable tasks.

A Different Approach to AI Innovation

While Chinese tech giants like Alibaba have hundreds of thousands of employees, DeepSeek operates more like a research lab, employing young graduates and PhD students from China’s top universities.

In his July interview, Liang suggested that China’s largest tech firms might struggle to lead the future of AI, citing their high costs and rigid corporate structures.

By contrast, DeepSeek’s lean operation and decentralized management may offer a more agile approach to AI development—one that continues to challenge both domestic and global competitors.