After DeepSeek, Alibaba Launches New AI Model, Claims It Surpasses OpenAI's GPT 4

Chinese tech major Alibaba on Wednesday announced the launch of a new AI model—Qwen 2.5-Max—which, it claims, surpasses the models of overseas rivals OpenAI and Meta Platforms.

Qwen 2.5-Max also outperforms DeepSeek-V3, the model unveiled recently by domestic rival DeepSeek, which has shook the global tech market as it matches the capabilities of OpenAI's ChatGPT and Meta's Lama-3.1, while being developed at a fraction of their costs.

"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 update posted on Chinese social media platform WeChat.

The new AI model has been launched on Alibaba Cloud Bailian platform, Alibaba reportedly said. It has surpassed DeepSeek-V3 on various benchmarks, including LiveBench, LiveCodeBench, MMLU-Pro and Arena-Hard, the company claimed, as per reports.

The launch of Alibaba's advanced AI model, in the aftermath of DeepSeek-V3, further raises question on the dominance of the US-based Big Tech firms, who are relying excessively on high-cost semiconductor chips.

Since the launch of DeepSeek-3, major tech firms in the US have cumulatively logged a plunge of over $1 trillion in valuations, with Nvidia Corp. contributing the most to the decline.

The rout was driven by the assumption that Wall Street majors, including Microsoft Corp., may not hold the edge in the next phase of AI, as companies with leaner development costs have launched models that are seen as more efficient.

While the latest developments have rattled the US tech market, analysts see some positives for India. DeepSeek's success sparks optimism as the Chinese startup has been able to challenge American tech firms despite the restrictions imposed by Washington on the supply of advanced AI chips, they point out.

The cost reductions demonstrated by DeepSeek have profound implications for global AI dynamics. "Training costs are lower, and I don't need Nvidia's higher-strength GPUs. AMD or Qualcomm GPUs could suffice,” Deepak Shenoy, chief executive officer of Capitalmind, told NDTV Profit. "This is driving fear in US markets, but it's very good news for India in the long term."