In a significant advancement for neuroscience and artificial intelligence, India has introduced MANAS-1, a groundbreaking AI model designed to interpret brain signals for the early detection of neurological and psychiatric disorders.
Developed by the deep-tech firm NeuroDx, this 400-million-parameter Brain Language Foundation Model represents a major step forward in using AI to understand the “language” of the brain, potentially transforming diagnostics and treatment in healthcare.
What is MANAS-1 and How Does It Work?
MANAS-1 is essentially an AI foundation model tailored for brain electrophysiology, functioning much like large language models such as ChatGPT but applied to electroencephalography (EEG) data. EEG measures the electrical activity produced by the brain through electrodes placed on the scalp, capturing patterns that reflect neural communication.
Unlike traditional imaging techniques like MRI, which provide structural snapshots, EEG reveals dynamic brain functions that can indicate subtle abnormalities before symptoms manifest. MANAS-1 processes these signals to “comprehend the fundamental language of the brain,” as described by Dr. Puneet Agarwal, a neurologist and former professor at the All India Institutes of Medical Sciences who led its development.
This allows the model to identify patterns associated with disorders, offering insights that go beyond what human experts or conventional tools can achieve alone.
The model’s architecture interprets brain functional connectivity and neural network dynamics, providing a unified framework for brain health screening. It has demonstrated over 95% accuracy in detecting biomarkers for epilepsy and other mental health disorders, marking a high level of reliability in preliminary tests.
As Dr. Agarwal explained, “MANAS-1 is designed to ‘understand the basic language of the brain.’” He further noted that it is a “foundation model similar in concept to ChatGPT that learns from large-scale EEG data to interpret brain signals that traditional tests like MRI cannot fully decode.”
The Data Behind the Model
At the core of MANAS-1 is an extensive dataset comprising 60,000 hours of brainwave recordings gathered from more than 25,000 patients. This vast collection of EEG data enables the AI to learn intricate patterns in brain activity, much like how language models are trained on text corpora to predict and generate responses.
The training process focuses on decoding electrical signals to reveal hidden insights into neurological conditions, ensuring the model can generalize across diverse patient profiles. This approach addresses a key challenge in neuroscience: the complexity and variability of brain signals, which often require specialized expertise to analyze.
Development and Launch
MANAS-1 was developed by NeuroDx under the guidance of Dr. Puneet Agarwal and founder Siddharth Panwar, with computational support from the Indian AI Mission under the Ministry of Electronics and Information Technology (MeitY). The project began with an announcement of its creation on February 17, 2026, and culminated in its official launch at the India AI Summit on February 23, 2026.
To foster collaboration, the model has been released as open source on Hugging Face, allowing researchers worldwide to build upon it. NeuroDx has also submitted MANAS-1 for government evaluation as a sovereign AI asset, which could unlock funding for further enhancements.
As Panwar stated, “The model aims to bring major changes to diagnostics, especially in neuroscience, by enabling early detection of brain disorders and more accurate treatment.” He added, “Huge opportunities are there in the AI sector. Our product will match highest standards.”
Dr. Agarwal highlighted the model’s significance in a LinkedIn post: “MANAS-1 represents a paradigm shift in artificial intelligence, enabling machine understanding of brain electrophysiology as a structured biological language. This foundation model enables scalable, accurate, and early detection of neurological and psychiatric disorders and advances the field of brain-computer interfaces and computational neuroscience.”
He further emphasized, “We believe MANAS-1 will play a transformative role in building AI systems that can truly understand and interpret the human brain.”
Applications and Potential Impact
MANAS-1 serves as a foundational platform for creating specialized AI tools targeting specific conditions like epilepsy, dementia, and other psychiatric disorders. It can facilitate preliminary screenings at community health centers, Ayushman Arogya Mandirs, and district hospitals, helping to bridge the gap caused by the shortage of neurologists and psychiatrists, particularly in rural India.
By enabling timely referrals, the model could reduce detection delays and long-term healthcare costs. Beyond diagnostics, it holds promise for advancing brain-computer interfaces, cognitive assessments, and precision neuroscience, making brain health checks more accessible and non-invasive.
However, any disease-specific applications derived from MANAS-1 will require rigorous regulatory evaluation and approvals before clinical deployment to ensure safety and efficacy. The model’s integration into healthcare systems could democratize access to advanced neurological care, especially in a country like India where specialist shortages are acute.
Looking ahead, NeuroDx plans to release MANAS-2 in the coming weeks and scale the model to 2 billion or even multi-billion parameters for deeper brain intelligence modeling. This evolution aims to enhance its capabilities in neurological diagnostics, neuroscience research, and next-generation technologies.
As part of broader efforts, it contributes to building India’s first foundation-scale physiological state model for interpreting human biosignals. With government support and open-source accessibility, MANAS-1 positions India at the forefront of AI-driven brain health innovations.
About the Author
