Kazakhstan-based startup Cybernet AI has unveiled the largest automatic speech recognition (ASR) model designed specifically for Turkic languages. The project marks a major technological milestone for Central Asia and represents one of the first full-scale AI models created from the ground up for an entire linguistic group, rather than adapted from English-language solutions, DKNews.kz reports.
The project was implemented with strategic support from Astana Hub and the Ministry of Artificial Intelligence and Digital Development of Kazakhstan, and trained using Microsoft’s GPU infrastructure.
Why this matters
Until now, Kazakhstan and other Central Asian countries largely relied on international ASR models primarily optimized for English. These solutions often struggled with Turkic phonetics, regional accents and the widespread use of mixed speech, limiting the quality of voice-based AI services in banking, telecommunications and public services.
Cybernet AI chose a different path – building a model inside the region and for the region, embedding the linguistic characteristics of modern Turkic languages from the start.
Six languages and real-life mixed speech
The new ASR model was trained on:
- Kazakh
- Turkish
- Uzbek
- Kyrgyz
- Azerbaijani
- Tatar
In addition, it accurately recognizes mixed Turkic-Russian speech, a common format in real customer interactions with businesses and government services.
The model demonstrates high resilience to background noise, accents and language switching within a single sentence, while maintaining strong accuracy across different use cases – from fintech to public services.
From technology consumer to technology creator
According to Tanat Uskembayev, Managing Director of Astana Hub, the project has strategic significance not only for Kazakhstan, but for the entire Turkic-speaking world.
“Cybernet AI’s development enables countries of the Turkic language group to scale digital services and customer support in their native languages using artificial intelligence without compromising quality and at an affordable cost. For Kazakhstan, this represents a shift from being a consumer of global AI solutions to becoming a developer of its own technologies that are competitive at the international level.”
AI that understands how people really speak
Cybernet AI’s Chief Technology Officer Rashid Ismailov emphasized that the company deliberately avoided adapting English-based models.
“We consciously rejected the adaptation of English-language models and created an ASR system that natively understands natural, mixed conversational speech. This makes the project not just another tech product, but a contribution to building a local AI ecosystem and strengthening Kazakhstan’s technological sovereignty.”
Designed for business and government
The ASR model is targeted at organizations handling large volumes of voice communications, including:
- banks and financial institutions
- telecom operators
- contact centers of all levels
- logistics and service companies
- government bodies providing public services
Developers say the solution can significantly reduce customer support costs while enabling 24/7 voice-based service, which remains the most in-demand communication channel.
About Cybernet AI
Cybernet AI is a resident of Astana Hub specializing in secure enterprise AI solutions for contact centers in banking, fintech and other regulated industries. Its platform focuses on high performance, controlled scalability and substantial reductions in operating costs, while meeting strict security and regulatory requirements.
The company operates in Kazakhstan, Uzbekistan, Kyrgyzstan, the United States, Mexico and Egypt, and is implementing projects in the UK and EU countries. Cybernet AI solutions support 34 languages, including Kazakh, Uzbek, Russian, English, Arabic and Spanish.
A new AI market for the Turkic world
Experts note that the launch of this ASR model could become a catalyst for the development of an independent AI market for Turkic languages, with its own technologies, talent base and export potential.
Projects like this are increasingly shaping who will set the rules in the global artificial intelligence landscape.
