Chinese tech company Alibaba has launched a new version of its Qwen AI model, claiming it surpasses the much-vaunted DeepSeek.
DeepSeek launched its way onto the scene, tanking the stock of major US companies like Nvidia and its connected partners with a large language model to compete with the giants, made for just a fraction of the cost. Now, Alibaba’s Qwen wants to make its own mark, released on the first day of the Chinese New Year.
“Qwen 2.5-Max outperforms … almost across the board GPT-4o, DeepSeek-V3 and Llama-3.1-405B,” Alibaba’s cloud unit said in an announcement posted on its official WeChat account – as reported by Reuters. The other models mentioned are OpenAI and Meta’s most advanced open-source AI models.
Alibaba enters an increasingly competitive marketBeing a holiday in China, the rush to release Qwen could be a response to the furore around DeepSeek. Indeed, other tech companies are scrambling to show off what they have to offer, with OpenAI CEO Sam Altman being quick to both compliment DeepSeek and assert that his company can still do better.
At the same time, TikTok owner ByteDance has released an update to its own AI model, claiming it outperformed OpenAI in a widely-used benchmark test, AIME.
It’s not the first time that DeepSeek has triggered chaos in the tech industry, with its preceding model, DeepSeek-V2, sparking a price war in China last May. DeepSeek has a habit of offering open-source and shockingly cheap models, with V2 costing only 1 yuan ($0.14) per 1 million tokens and V3 launching after less than $6 million worth of investment.
Back in July, DeepSeek founder Liang Wenfeng maintained that his startup “did not care” about price wars in an interview with Chinese media publication. Instead, the company’s focus is on achieving AGI (artificial general intelligence). That in itself is a contrast to some American AI founders, with Altman looking at AI superintelligence.
Featured image: Midjourney
The post Alibaba claims new AI model surpasses newcomer DeepSeek appeared first on ReadWrite.