African companies combine Chinese and Western AI models
Companies in Africa are combining Chinese and Western artificial intelligence models, choosing them depending on cost, performance and the specific task. As the South China Morning Post reports, Nigerian insurance technology platform Curacel added Zhipu AI’s Chinese GLM-5.3 model to its infrastructure alongside models from Western providers.
Curacel CEO and co-founder Henry Mascot said leading Chinese models show comparable results at a lower cost in high-volume tasks. These include programming, data extraction, classification and customer support. At the same time, in his assessment, Western systems remain stronger in the most complex tasks that require reasoning and high reliability.
Cost and local languages
Nairobi-based consulting company EverseTech moved some clients to Chinese models because of the lower price per million tokens. According to the company’s co-founder Michael Michie, a client in the legal sector changed models because of costs, while another chose a Chinese model after it converted documents into numerical representations for retrieval in a retrieval-augmented system faster and more cheaply.
In Uganda, developers of the Sunflower model, which provides farmers with weather updates and agricultural advice in more than 30 local languages, used Qwen3. They found that this model works better with Ugandan languages than models from Meta Platforms or Google. Ivorian fintech company Djamo also allocates models depending on task complexity: lighter Chinese models are used for simple customer queries, while more powerful ones are used for writing or reviewing code.
More current news is available on the UA.News Telegram channel Telegram.
Open weights and dependence on providers
Some Chinese developers provide model weights for download. This allows companies to adapt models and, if they have sufficient computing resources, run them on their own infrastructure. Sanifu AI co-founder Bernard Momani Nyagaka noted that this approach could help banks and fintech companies keep sensitive data, including banking records and KYC documents, within their own systems.
At the same time, local deployment requires computing capacity and technical expertise. Curacel uses a single interface that can route tasks to Chinese, American, European or locally deployed models, reducing dependence on a single provider.
The growing use of Chinese models is taking place amid U.S.-China competition in AI, which also encompasses semiconductors, cloud infrastructure, critical minerals and technology regulation rules. A draft U.S. State Department cable reviewed by Reuters on August 14 warned that states joining the China-created World Artificial Intelligence Cooperation Organization could be excluded from the U.S.-led Pax Silica coalition.
