The Kazakhstan-based startup Cybernet AI has unveiled the largest Automatic Speech Recognition (ASR) model specifically designed for Turkic languages, according to the Ministry of Artificial Intelligence and Digital Development of the Republic of Kazakhstan.
This project marks a milestone as the first large-scale AI development of its kind to emerge from Central Asia. Unlike previous solutions that were adapted from English-centric frameworks, this model was built from the ground up to address the linguistic nuances of the Turkic language group.
The new model supports Kazakh, Turkish, Uzbek, Kyrgyz, Azerbaijani, and Tatar. Crucially, it is engineered to accurately process “code-switching”—the mixture of Turkic and Russian speech—a format prevalent in daily life, business interactions, and public service delivery across the region.
Previously, Kazakhstan and its neighbors relied on international ASR models optimized for English. These global solutions often struggled with Turkic phonetics, local accents, and mixed-language speech, which limited the effectiveness of voice-based AI services. Cybernet AI has bypassed these limitations by developing a localized architecture that captures the linguistic DNA of the region.
The new ASR model is primarily targeted at sectors managing high-volume voice communications, including: banking and finance, telecommunications, logistics and service companies, government agencies.
Developers emphasize that the technology will significantly reduce operational costs in customer support channels while ensuring 24/7 accessibility. The model has proven resilient against background noise and varied accents, maintaining high accuracy across diverse sectors from fintech to social services.
The project was realized with strategic support from Astana Hub and the Ministry of Artificial Intelligence and Digital Development, utilizing Microsoft’s GPU infrastructure.
About Cybernet AI: An Astana Hub member, Cybernet AI specializes in secure corporate AI solutions for regulated industries. While rooted in Central Asia, the company maintains a global footprint with operations in Kazakhstan, Uzbekistan, Kyrgyzstan, the USA, Mexico, and Egypt, as well as ongoing projects in the UK and the EU. Their platforms currently support 34 languages, including Arabic, Spanish, and English. ///nCa, 20 January 2026