China’s tech giant Alibaba unveiled its latest Qwen 2.5 AI model on Wednesday, claiming its superiority over the widely praised DeepSeek-V3.
The announcement came through Alibaba’s cloud division, which declared via WeChat that the new Qwen 2.5-Max model outperforms not only DeepSeek-V3 but also OpenAI’s GPT-4o and Meta’s Llama-3.1-405B in benchmarks.
The timing of this release speaks volumes about the competitive pressure within China’s AI sector. Launching on the first day of Lunar New Year celebrations, when most Chinese citizens are traditionally at home with family, shows the urgency Alibaba feels in responding to DeepSeek’s remarkable three-week rise to fame.
The generative AI industry has seen a seismic shift following DeepSeek’s back-to-back releases – its AI assistant powered by DeepSeek-V3 on January 10, followed by the R1 model on January 20.
These launches shook up Silicon Valley greatly, triggering a significant downturn in technology stocks as investors began questioning the massive expenditure plans of leading US. AI companies in light of DeepSeek’s reportedly minimal development and operational costs. ByteDance, the company behind TikTok, unveiled an upgraded version of its flagship AI model merely two days after DeepSeek-R1’s debut.
ByteDance’s announcement claimed their updated model surpassed Microsoft-backed OpenAI’s o1 in the AIME benchmark test, a crucial measure of AI systems’ ability to comprehend and execute complex instructions. DeepSeek’s decision to open-source its V2 model and offer processing at just 1 yuan ($0.14) per million tokens also forced Alibaba Cloud to respond with sweeping price reductions across its model lineup, slashing rates by as much as 97%. This could signal a new era of accessibility in the AI market, hopefully outside of China as well.