Islamabad, Jan 29: Chinese tech giant Alibaba unveiled its latest AI model, Qwen 2.5, claiming it surpasses its competitors, including DeepSeek-V3, OpenAI’s GPT-4, and Meta’s Llama-3.1-405B. This announcement was made by Alibaba’s cloud division on WeChat, with the company asserting that Qwen 2.5-Max outperformed these models in benchmark tests.
The release timing is significant, coinciding with the Lunar New Year celebrations, a period when most Chinese citizens are traditionally at home with family. This suggests Alibaba’s urgency in responding to the DeepSeek model’s growing success. DeepSeek made waves with its AI assistant powered by DeepSeek-V3 on January 10, followed by the launch of its R1 model on January 20, which garnered significant attention in the tech world.
Read More:China’s DeepSeek AI: Disrupting Global Tech Race | Analysis
The rapid rise of DeepSeek’s models has sparked a shift in the generative AI industry, triggering a downturn in Silicon Valley tech stocks. This is largely due to the low operational and development costs claimed by DeepSeek, which contrasts with the massive investments made by leading U.S. AI companies. Additionally, ByteDance, the parent company of TikTok, introduced an upgraded AI model just two days after DeepSeek’s R1 model, claiming it surpassed OpenAI’s o1 in the AIME benchmark test. This test measures an AI’s ability to understand and execute complex instructions.
In response to DeepSeek’s open-sourcing of its V2 model and its low-cost processing priced at just 1 yuan ($0.14) per million tokens, Alibaba Cloud significantly reduced the prices of its own AI models. Some price reductions were as steep as 97%, which could herald a new era of affordable and accessible AI technology potentially extending beyond China as well.
