On Wednesday, Chinese tech giant Alibaba launched Qwen2.5-Max, a new version of its AI model that it claims outperforms DeepSeek-V3, a highly praised model.
The timing of the Qwen2.5-Max release, coinciding with the first day of the Lunar New Year when most Chinese workers are on holiday, highlights the pressure from DeepSeek’s rapid rise over the past three weeks. This surge has impacted not only international competitors but also stirred activity within China’s domestic tech scene.
Alibaba’s cloud division, in an announcement posted on its official WeChat account, stated that “Qwen 2.5-Max outperforms … almost across the board GPT-4, DeepSeek-V3, and Llama-3.1-405B,” referring to OpenAI and Meta’s leading open-source AI models.
DeepSeek’s AI assistant, powered by its DeepSeek-V3 model, launched on January 10, followed by the R1 model release on January 20. These releases have shaken Silicon Valley, causing a dip in tech stocks. DeepSeek’s reportedly low development and usage costs have raised questions about the massive spending plans of major U.S. AI firms.
This success has spurred a rush among Chinese competitors to enhance their own AI models. Just two days after DeepSeek’s R1 release, ByteDance, the parent company of TikTok, rolled out an update to its flagship AI model, claiming it outperformed Microsoft-backed OpenAI’s o1 in AIME, a benchmark for testing AI’s understanding of complex instructions. This mirrors DeepSeek’s own claims that its R1 model competes with OpenAI’s o1 on several performance benchmarks.
Last May, the release of DeepSeek’s V2 model ignited a price war in China’s AI sector. Its open-source nature and extremely low cost of just 1 yuan ($0.14) per 1 million tokens, or data processing units, prompted Alibaba to slash prices by up to 97% across a range of its models.
Other Chinese tech companies, including Baidu and Tencent, followed suit. Baidu launched China’s first ChatGPT-like model in March 2023, while Tencent, the country’s most valuable internet company, also joined the competition.
In a rare interview with Waves, DeepSeek’s elusive founder Liang Wenfeng explained that the startup was not concerned with price wars, emphasizing that their focus was on achieving Artificial General Intelligence (AGI). OpenAI defines AGI as autonomous systems that exceed human performance in most economically valuable tasks.
While large Chinese tech firms like Alibaba employ hundreds of thousands of people, DeepSeek operates more like a research lab, primarily staffed by young graduates and PhD students from China’s top universities. Liang, in the same interview, pointed out that he believed the traditional structures of large tech companies might not be ideal for the future of AI, highlighting the efficiency and flexibility of DeepSeek’s lean, loosely managed operation.
