African startups turn to Chinese open-weight AI models to cut costs

In short: Developers in several African tech hubs are increasingly using Chinese open-weight AI models because they cost less and can run locally.

What's going on

Developers and startups in countries including Nigeria, Kenya, South Africa, and Rwanda are adopting Chinese AI models such as DeepSeek and Alibaba’s Qwen. These models are showing up more often in new products like study helpers, customer support bots, and local-language assistants.

A big reason is price. The reporting cited costs of about $0.27 per million input tokens and $1.10 per million output tokens for some DeepSeek models, compared with about $5 and $15 for OpenAI’s GPT-4o. A token is a small chunk of text (think of it like a few characters or part of a word), and apps can use millions of them quickly.

Another factor is that many of these Chinese models are “open-weight,” meaning developers can download the model and run it on their own computers or servers (like owning a copy of software instead of renting it). That matters in places where internet connections can be unreliable or expensive, since relying on US cloud services often means every request must travel overseas and be billed in US dollars.

What to watch

Top US models still tend to do better on many performance tests, so big banks and large companies may keep paying for them. For startups and small teams, the key question is whether “good enough” models that run locally will keep improving, and how African governments and companies handle concerns about dependence on any one foreign supplier.

Source: NYTimes