By Umar Abakirov
DUSHANBE, Tajikistan, July 5. The first AI
model built specifically for the Tajik language and its regional
dialects has been officially unveiled, Trend reports via its developers
zehnlab.ai.
The new model, called Soro, marks a significant step forward in
making artificial intelligence more inclusive for underrepresented
According to the developers, Tajik has historically been
underrepresented in the training data of major large language
models such as GPT and Llama. Soro was created to change that by
focusing on the linguistic richness of Tajik — including its
non-standard syntax, dialectical variations, and unique patterns of
real-life usage.
The team behind Soro says the model is designed not just to
answer simple queries, but to understand and reflect the cultural
and linguistic diversity of the people of Tajikistan. With its
multimodal capabilities under development, Soro aims to bridge the
gap between technology and local language communities.
Meanwhile, on June 25, President of Tajikistan Emomali Rahmon
inaugurated the Darya AI Artificial Intelligence Management Center
in the Darvoz district. The center was established as part of the
national initiative declaring 2025–2030 the Years of Development of
the Digital Economy and Innovation.