As China welcomed the Year of the Snake, Alibaba uncoiled and struck; unleashing its latest AI model, Qwen 2.5-Max.
In a stinging shot at its closest rivals, Alibaba’s cloud unit declared, “Qwen 2.5-Max outperforms…almost across the board, [OpenAI’s] GPT-4o, DeepSeek-V3 and [Meta’s] Llama-3.1-405B.”
This is an unusually direct challenge to fellow AI vendors.
Tech analyst Steve Nouri highlighted the model’s potent capabilities, and questioned the tech industry’s hardware assumptions:
What if we really don’t need those many advanced GPUs? Even Anthropic’s CEO just said they have only spent a few $10 millions to train their models…With 20 trillion tokens of training data and advanced RLHF tuning, this model is setting new standards for AI intelligence.
Alibaba’s new engine contains 20 trillion tokens of training data and has RLHF tuning; pushing the boundaries of AI performance, and confronting tough topics head-on, striving to always provide direct answers to users’ questions.
The Upstart Challenge
DeepSeek, a 20-month-old startup from the same Hangzhoe region as Alibaba, has hypnotised Silicon Valley with its lean approach. Operating more like a research lab than a traditional tech giant, their team of young graduates and doctorate students triggered an industry upheaval this week. Their entry into the market at just 1 yuan ($0.14) per million tokens forced a fundamental rethink of AI pricing models, culminating in yesterday’s dramatic GenAI release from Alibaba.
The timing of Qwen 2.5-Max’s launch – as China brings in its new year and businesses typically pause for family celebrations – underscores Alibaba’s urgency to establish itself as a major player in the generative AI race.
An arguably forced move; it comes after Alibaba’s cloud unit had already announced drastic price cuts of up to 97% on their AI models, with Baidu and Tencent following suit. The unusual holiday release suggests Alibaba is determined to reclaim its position at the forefront of Chinese AI development.
Wall Street Wobbles
Alibaba’s stock rose 3% to early November highs following its Lunar New Year AI announcement, but the news rippled through the broader tech sector against a backdrop of existing market anxiety.
The previous few days had already seen NVIDIA stocks twist violently – shedding $590 billion in market value Monday, before recovering $260 billion Tuesday, only to drop another $130 billion on Wednesday. Alibaba’s announcement didn’t have the same seismic impact that DeepSeek did, but there’s no telling what’s around the corner in the fast-paced world of GenAI.
Industry Evolution
The whirlwind week we’ve experienced is likely to shape the AI industry’s future, as traditional players shed their rigid structures and look to become more nimble. DeepSeek’s founder, Liang Wenfeng, believes established tech giants may currently be too constricted by their own size to adapt quickly in AI innovation.
For businesses worldwide, this intense competition could make advanced AI tools more accessible, reducing reliance on costly infrastructure. As the Year of the Snake unfolds, it may mark a turning point where efficiency and innovation slither past sheer computational power and capital.
