Pakistani tech firm Data Vault has introduced the country’s first locally developed generative pre-trained transformer (GPT) platform, ‘Zahanat AI,’ in a milestone development for Pakistan’s AI sector. The platform, launched on Thursday, is designed to cater specifically to local needs by integrating cultural and linguistic elements unique to Pakistan.
Zahanat AI is a text-based generative AI model capable of engaging in human-like conversations, answering user queries, and assisting in various domains. According to Data Vault, the platform has been trained on local datasets, making it the first AI model tailored to Pakistan’s regional challenges. “The problems and solutions it provides will align with our culture and environment,” said Syed Muhammad Yahya, Chief Technology Officer (CTO) of Data Vault, during the AI model’s launch in Islamabad.
Generative AI models like Zahanat AI are trained on large datasets to generate human-like text, images, and other content. Similar to international models such as ChatGPT and DeepSeek, Zahanat AI supports English, Urdu, and local languages, with plans for further expansion. Unlike foreign AI platforms, however, Zahanat AI is hosted entirely within Pakistan, a factor that could enhance data security and regulatory compliance.
Privately funded through friends and family, the AI model is built on Meta’s Large Language Model Architecture (LLaMA), using 1.5 billion parameters. Yahya explained that the development process took six to eight months and required a multi-GPU architecture involving a stack of 12-15 GPUs. The GPT platform is currently available in a limited capacity within Pakistan, with plans to launch a mobile application soon.
The unveiling of Zahanat AI coincides with Pakistan’s efforts to finalize its first AI policy, which aims to foster AI-driven innovation through public-private partnerships. Dr. Aneel Salman, who is overseeing the upcoming National Artificial Intelligence Policy, emphasized that the policy is designed to encourage local AI development.
“Our AI policy is very liberal. It promotes the local ecosystem and supports local entrepreneurs.”
One of the primary motivations behind local AI development is cybersecurity. Dr. Salman warned that foreign AI models pose potential risks, as user data uploaded to these platforms becomes part of their algorithms. “You don’t know how that document will later be deciphered,” he said, underscoring the importance of self-reliance in AI technology.
During the launch event, the developers demonstrated Zahanat AI’s capabilities by answering questions related to healthcare and mathematics. However, the model faced challenges when responding to a query about freelancing expert Hisham Sarwar, initially lacking details and later providing inaccuracies. Dr. Salman acknowledged that such limitations are expected in early versions of AI models.
“When ChatGPT was first launched, it had problems. DeepSeek also had issues in its initial phases. Zahanat AI will improve over time with versions like Z1, Z2, and Z3.”
Experts believe that the long-term success of Zahanat AI will depend on its adoption by the public and integration with research and development initiatives. Prof. Dr. Yasar Ayaz, chairman of the National Center of Artificial Intelligence (NCAI), highlighted the need for collaboration between AI companies and local innovators.
“If we want to prevent initiatives like Zahanat AI from failing, we must use their technology more extensively. They need to connect with innovators in R&D setups across Pakistan to jointly develop groundbreaking solutions.”
As Pakistan ventures into AI development, Zahanat AI represents an important step toward self-sufficiency in artificial intelligence. While the model still requires refinements, it has the potential to revolutionize Pakistan’s digital landscape, offering localized AI solutions that address the country’s unique needs.
