Hoosh Afarin Large Language Model
Hoosh Afarin is a Persian Large Language Model (LLM) fine-tuned using 45 billion Persian tokens to optimize performance across specialized domains such as legal services, religious studies, education, contact center support, and judicial applications. The model has successfully completed all stages of Instruction Tuning and Task Tuning and is now capable of performing advanced Agent and Agentic tasks within these domains.
One of Hoosh Afarin’s key features is its integrated Guard Layer, designed to prevent inappropriate prompts and unsafe responses while ensuring secure and reliable usage. With high accuracy, the model can provide legal analysis, support educational processes, enhance customer service operations, and assist in judicial and regulatory workflows.
Hoosh Afarin represents a significant step toward the development of practical, sovereign, and domain-specific AI technologies tailored to the needs of Persian-speaking communities.
Founding Story
The Vice Presidency for Science and Technology launched a national initiative to develop a Persian-native LLM with localized linguistic and operational capabilities. Four companies participated in the project through a collaborative consortium model.
Following the successful completion of the initiative, three members of the consortium introduced the official commercial release of Hoosh Afarin, delivering one of the first
Target Industries
Industrial and mining AI service operators and adopters
- Service Delivery Model: On-Prem Deployment and Cloud Services
- Export-Oriented Product Vision: No
- Smart Money Compatibility: Yes
- Activity Launch: 2021
- Hoomas Investment Date: 2024
Progress & Roadmap
Development of the “Kalameh” crisis-management chatbot system with Iran-access operation capability during internet outages, supporting emergency response, firefighting services, law enforcement, psychological support, and crisis management during military threats
Development of industrial crisis-management systems for operation during military and security threats
Deployment of conversational knowledge-management systems for Predictive Maintenance (PdM) applications