Alibaba upgrades AI model Tongyi Qianwen, releases industry-specific models

CHINESE technology giant Alibaba said on Tuesday (Oct 31) that it has updated its artificial intelligence (AI) model Tongyi Qianwen and released a suite of industry-specific AI models amid an intensifying AI race among tech companies. Alibaba’s cloud computing arm at its annual conference in Hangzhou said Tongyi Qianwen 2.0 has “hundreds of billions of” parameters – a benchmark used to measure AI model power – making it one of the world’s most powerful AI models by that metric.

It also said it has launched eight AI models for the entertainment, finance, healthcare and legal industries.

The upgrade comes just six months after the model’s initial release, reflecting the pace at which tech companies are racing to control China’s nascent and quickly expanding AI market.

Rival Tencent last month said China was in the middle of a “war of a hundred models”, with over 130 flooding the market. Tencent said its Hunyuan AI has more than 100 billion parameters and exceeds OpenAI’s GPT-4 model in processing Chinese.

The GPT-3 AI model from Microsoft-backed OpenAI contained 175 billion parameters in 2020 and Meta Platform’s Llama 2 model had 70 billion parameters in 2023.

Alibaba’s industry-specific models went online on Tuesday, offering dedicated tools for creating images, writing computer code, analysing financial data and searching for legal documents.

SEE ALSO

Alibaba promises major discounts ahead of China’s ‘Singles Day’ shopping event

Alibaba, Tencent join funding for Chinese AI high-flyer Baichuan

Belgium’s intelligence service monitors Alibaba hub in Europe over ‘espionage’ concerns: FT

Asean Intelligence

Get insights into businesses across South-east Asia

Get the free report

At the conference, chairman Joe Tsai also said about half of large-language AI models in China now run on Alibaba Cloud.

He said Alibaba’s AI model-sharing platform ModelScope now features 2,300 models and 2.7 million contributing developers. REUTERS