Pakistan’s first locally built GPT AI ‘Zahanat AI’ launched

Pakistan is also drafting its first AI policy to encourage domestic AI development
An undated image. — Canva
An undated image. — Canva

Pakistan has introduced its first homegrown generative AI model, ‘Zahanat AI,’ developed by tech firm Data Vault. The GPT platform aims to provide localised AI solutions, supporting English, Urdu, and regional languages.

Zahanat AI is a text-based generative AI model that enables human-like conversations, answers queries and assists in areas like healthcare, mathematics, and coding.

Pakistan’s first GPT AI ‘Zahanat AI’

Unlike foreign AI models, it is trained on Pakistani culture and local challenges. “The problems and solutions it provides are based on our culture and environment,” said Data Vault CTO Syed Muhammad Yahya at the launch event in Islamabad.

Large data sets are utilized by generative AI models such as ChatGPT, DeepSeek, and Meta-AI to create text. Zahanat AI is based on Meta's LLaMA architecture with 1.5 billion parameters.

Moreover, the development was completed in six to eight months using a multi-GPU system with 12-15 GPUs. The model is now being made available for limited access within Pakistan, and a mobile application is being developed.

Pakistan is also drafting its first AI policy to encourage domestic AI development. "Our AI policy incentivises local innovation and entrepreneurs,” said Dr Aneel Salman, who is leading the policy initiative. He also highlighted cybersecurity risks associated with foreign AI models, stressing the need for local alternatives.

Experts believe public adoption is crucial for Zahanat AI’s success. “If we want this initiative to grow, we must actively use local AI technology,” said the Chairman of NCAI Prof. Dr Yasar Ayaz.