HEFEI, Aug. 11 (Xinhua) -- Inside the smart exhibition hall of Safaricom, Kenya's largest telecommunications operator, a digital human supported by a Chinese tech company delivers live demonstrations of core business services and interprets financial report data for visitors through intelligent interactive functions.
Across emerging markets with huge digital-economy potential, Chinese AI companies are partnering with host nations to build home-grown large language models adapted to local languages and industrial demands.
Leading global large language models have long prioritized widely spoken languages, leaving minor-language communities underserved. Multiple experts warn that countries with less-used languages run real risks of being sidelined by AI, and this technological gap has become a key focus for Chinese AI firms.
Leveraging its expertise in speech intelligence and natural-language processing, Chinese tech company iFLYTEK has ramped up research and development for multilingual technologies. Its speech large language model Spark supports speech recognition and simultaneous interpretation for more than 130 languages.
With the ASEAN market in mind, iFLYTEK has forged deep cooperation with south China's Guangxi Zhuang Autonomous Region. Drawing on this region's pool of language talent for ASEAN countries, rich language corpora and digital infrastructure, the company has rolled out the Spark ASEAN Multilingual Large Model Base.
The base covers 10 languages including Malay, Indonesian, Vietnamese and Thai. It delivers performance comparable to top international models with smaller parameter sizes, enabling efficient and secure deployment of industry-specific AI applications across ASEAN member states.
The company has signed cooperation deals with partners in countries such as Laos, Malaysia and Thailand. In Malaysia, it has jointly set up an AI multilingual intelligent dubbing and translation center with Enjoy TV & Film Broadcasting Corporation. Powered by the model, this facility supports translation for over 130 languages and aims to foster an Asian-oriented intelligent dubbing ecosystem.
"Mainstream English large models show clear shortcomings when serving many overseas markets. There is strong, genuine demand for large models customized for local tongues," said Dong Bin, vice president of iFLYTEK's brand marketing center.
Dong added that the company has also open-sourced its large model, allowing nearly 600,000 overseas developer teams to build custom products via its algorithms and speech tools. The firm is pushing for overseas roll-out of proven hardware-and-software AI solutions for education, healthcare and office scenarios.
More Chinese companies are offering diverse AI options for developing countries through varied business approaches.
Another example is DeepSeek, whose open-source model offers a more affordable option for African developers. At roughly 0.27 U.S. dollars per million input tokens and 1.10 U.S. dollars per million output, it is significantly cheaper than many alternatives. Its open-source nature further lowers the financial barriers for local developers to get started.
In Cairo, Huawei has launched its first 100-billion-parameter Arabic large model. Trained on local Arabic datasets from finance, power and oil-and-gas sectors, the model achieves 96 percent speech-recognition accuracy and serves more than 20 Arabic-speaking countries and regions. It helps these economies strengthen domestic AI capacity using locally sourced data that reflects their cultural and linguistic context.
Beijing-based company 01.AI, has jointly established Q.AI with Kazakhstan to develop and deploy large language models, AI agents and enterprise AI platforms tailored to local requirements.
"We respect local data sovereignty and prioritize on-site deployment and local operation under an open, collaborative and sustainable framework," said Kai-Fu Lee, CEO of 01.AI. He added that emerging economies like Kazakhstan are not advised to simply purchase models or outsource AI capabilities. Instead, they need home-grown, self-sustaining national AI capacity built around their own data, languages, industries and governance rules.
Chinese AI companies share a clear mindset for global outreach, which is to empower local partners rather than replace them. Their work seeks inclusive digital growth and helps developing countries narrow the digital divide -- a vision echoed by both Chinese enterprises and policymakers.
The 2026 World AI Conference took place in east China's Shanghai in July. At the event, China advocated developing AI in an open, win-win manner and helping Global South countries with AI capacity building. China has also pledged to provide developing countries with 5,000 opportunities in AI training and seminar programs over the next five years.
"China encourages open-source AI development, and its open-source large models are among the most-downloaded globally and underpin countless applications worldwide. Beyond that, China focuses on real support, to nurture local expertise so each country can develop AI applications suited to its own conditions," said Xue Lan, dean of the Schwarzman College at Tsinghua University in Beijing.