Alibaba, the Chinese tech behemoth, has developed Qwen 2.5, an updated AI model that it says surpasses top open-source models such as DeepSeek-V3. “Qwen 2.5-Max outperforms…” Alibaba’s cloud unit announced on WeChat that GPT-4o, DeepSeek-V3, and Llama-3.1-405B are used practically universally. The publication coincided with the Lunar New Year celebrations.
The announcement comes as DeepSeek’s aggressive pricing fuels growing rivalry in China’s artificial intelligence market. Its DeepSeek-V2 model, which debuted in May 2024, compelled big firms such as ByteDance, Tencent, Baidu, and Alibaba to cut prices.
DeepSeek’s AI helper and R1 model, released on January 10 and 20, respectively, have startled Silicon Valley. “Investors are questioning the huge spending plans of leading AI firms in the United States,” an analyst told Reuters. DeepSeek, founded by Liang Wenfeng in May 2023, uses a novel AI training approach and a cost-effective Mixture-of-Experts (MoE) architecture.
Its inexpensive API pricing—$0.55 per million input tokens versus OpenAI’s $15—has rocked the market.
Despite its low cost, DeepSeek lacks the powerful AI-generated media tools available in competing models. However, its effectiveness raises concerns about the long-term viability of the vast AI expenditures made by US computer corporations.
