Alibaba launches a super-sized AI model targeting ChatGPT

Alibaba launches a super-sized AI model targeting ChatGPT is profiled by BTW Media because published evidence links it to internet infrastructure, governance, operational dependencies, or market visibility.

• Alibaba, the Chinese tech giant, boasted that its super-sized artificial intelligence model outperforms OpenAI’s GPT-4 from US.
• According to Alibaba Cloud, they have announced that version 2.5 has shown various performance improvements compared to version 2.1 of Qwen SQL.
• Alibaba Cloud currently services Qwen models over 90,000 enterprises.

Alibaba Cloud announced its super-sized model, ‘Qwen 2.5’, and claimed that its performance in various aspects surpassed that of GPT-4 according to Chinese media outlet Diyi Caijing on May 9.

Surpasses ChatGPT

As China intensifies its push for technological self-reliance, the country’s leading tech giant Alibaba has unveiled a new artificial intelligence model. Alibaba officially launched the Tongyi Qianwen 2.5 on that day according to major foreign media outlets such as CNBC on May 9.

According to Alibaba Cloud, their Tongyi mega model is already being serviced by over 90,000 enterprises. Through their chat service DingTalk, the cumulative downloads of open-source models have exceeded 7 million from over 2.2 million enterprises. These models are being applied not only in PCs and mobile phones but also in fields such as automotive, aviation, astronomy, mining, education, healthcare, food service, and gaming.

Also read: HKBN, the first Alibaba Cloud partner certified in HK

Alibaba’s growth

Alibaba has officially announced the release of the enterprise version of its Tonyi Split Code. Tongyi Split Code, an intelligent coding assistant, has surpassed 3.5 million downloads since its plugin release in October 2023 and developers have adopted over 100 million lines of code, the company stated.

Also read: Alibaba sheds $360M Bilibili stake for AI push and business overhaul

Alibaba Cloud also announced the release of the latest open-source mega model, Qwen 1.5-110B, with 1.1 trillion parameters. They claimed superiority over Meta’s Lama-3-70B model in tests such as MMLU, TheoremQA, and GPQA. It was also highlighted that they secured the top position in the Open Large Language Model Leaderboard released by Hugging Face.