There’s been an escalation in the generative AI large language model “wars” as Alibaba Qwen 2.5 launched Wednesday. This latest AI salvo from China-based Alibaba is directly aimed at its in-country rival DeepSeek, which launched its own AI–DeepSeek-V3–in December 2024 and its R1 version in mid-January.
What sets DeepSeek-V3 apart from the other foundation AI models such as Claude, ChatGPT, Gemini, Llama and Perplexity is that its unique design came online much faster than the dominant players and required much less computing power to train compared to the other systems.
Why Alibaba Qwen 2.5 Launched
Because of its upgraded algorithm architecture, the V3 model reportedly produces comparable results as existing LLMs; however, the company states that it was able to train DeepSeek-V3 for less than $6 million using older Nvidia H800 GPU chips that debuted almost two years ago–that’s almost a lifetime within tech circles. Shortly after its release on January 20th, the DeepSeek-R1 AI assistant–powered by V3–became the top download within Apple’s Top Free App category.
On Tuesday, the reality of this achievement sunk in on Wall Street as investors sold off nearly $600 billion worth of Nvidia stock–questioning whether pricey next-gen GPUs such as its H200 and Blackwell processors will even be necessary. While Meta has reportedly scrambled to establish Llama war rooms to try and reverse engineer how the latest DeepSeek rollouts debuted so fast and cheap.
Alibaba Qwen 2.5 Versus DeepSeek-V3
But the premiere of DeepSeek’s latest innovations didn’t just take U.S.-based AI developers and chip makers off guard. Media outlets suggest that it spurred an AI upgrade by TikTok owner ByteDance and this latest AI launch of Alibaba’s Qwen 2.5. It’s reported that Alibaba specifically called out DeepSeek in a WeChat post stating that Qwen 2.5 “outperforms” V3.
While it’s too early to tell, which AI model from China will come out on top there are concerns surfacing about potential risks for both platforms. Issues that plagued China-owned TikTok are being raised regarding Qwen and DeepSeek-V3 regarding data security, privacy, potential misreporting of performance stats and separate issues of possible intellectual property theft on behalf of OpenAI and Microsoft–which could call into question whether V3 was trained from scratch or leveraged other AI models.
