Top AI Players Are Consolidating

Alibaba Group Holding's cloud computing business is reportedly finalizing the details of its collaboration with Chinese start-up 01.AI to develop artificial intelligence (AI) models. This is considered as a sign of consolidation in an increasingly crowded market on the mainland.

Alibaba Cloud announced last 2 January 2025 that it had reached a deal with Beijing-based 01.AI, founded by former Google China head Lee Kai-fu, to set up an "industrial large model laboratory".

The lab will combine the research and development capabilities of the two teams to explore technology and services related to AI models, with the aim to create "strong and comprehensive large model solutions" for business clients, according to a statement from Alibaba Cloud.

Founded in May 2023, 01.AI achieved unicorn status just six months later, after a funding round that included Alibaba Cloud. The start-up's open-source foundational large language model (LLM) Yi-34B was once ranked as the best pre-trained model by online AI community Hugging Face. It has since fallen behind.

01.AI's best model - Yi-1.5-9B-Chat-16K - came in 10th among Chinese competitors, according to the latest ranking by benchmarking platform SuperClue. Alibaba Cloud's Qwen2.5-72B-Instruct took the lead, trailing only models from foreign rivals OpenAI and Anthropic.

Lee, who previously worked at Microsoft and Apple, said 01.AI's joint lab with Alibaba Cloud will raise the bar "of global cutting-edge large models in terms of technology innovation".

The cooperation between two of China's top AI players comes as price wars continue in the domestic market, forcing companies to further slash prices or seek partnerships with former foes.

Alibaba Cloud said last 31 December that it would reduce the fees for using its visual reasoning AI model by up to 85 per cent, the third time it had marked down the prices of its AI services in the past year.

That came after TikTok parent ByteDance last month cut the price of its visual model to 0.003 yuan (US$ 0.0004) per thousand token uses, about 85 per cent lower than the industry average.