SoroLLM

SoroLLM is a Tajik-language large language model developed by zypl.ai as part of Project Soro — a national AI-in-education initiative covering 4,000+ schools across Tajikistan.

Overview

SoroLLM is designed to support educational content generation, question answering, and conversational interaction in Tajik, Russian, and English. It is the first foundation model optimized for the Tajik language.

🔗 Product website: sorollm.ai

Benchmarks

Benchmark	SoroLLM	GPT-4	Llama 3
Tajik MMLU (custom)	—	—	—
Tajik Translation (BLEU)	—	—	—

Заполните числами из ваших бенчмарков

Intended Use

• Educational content generation in Tajik language
• Student Q&A and tutoring assistance
• Tajik ↔ Russian ↔ English translation
• Integration with national LMS platform (e-donish)

Limitations

This model is optimized for educational contexts in Tajik and may not generalize well to other domains or languages outside its training data.

Contact

For partnership and access inquiries: zypl.ai