The year 2025 was a game-changer for AI in medicine and health care, marked by a sharp acceleration in drug discovery, diagnostics, genomics, and clinical workflows.
AI systems began outperforming humans in complex diagnostic tasks. Ambient scribes and clinical copilots significantly reduced clinician burnout and documentation time. It was also a year of AI-powered advances in precision medicine, regenerative tissues, and AI-designed antibiotics. India, too, made notable contributions to AI-driven medical science this year.
Here are some key AI models released in 2025:
Boltz-2 AI: Developed by MIT and Recursion Pharmaceuticals, this model enables ultra-fast prediction of protein–ligand complex structures and binding affinity (drug–target strength). The open-source model has accelerated virtual screening in drug discovery, potentially compressing development timelines significantly.
For your daily dose of medical news and updates, visit: HEALTH
Evo 2: A genomic foundation model trained on data from over 128,000 genomes spanning more than one lakh species across all domains of life, Evo 2 is described as a “representative snapshot of all known living species.” This open-source model can generate and predict novel DNA, RNA, and protein sequences, greatly enhancing the potential to engineer treatments, including for rare genetic conditions.
Pathology Bias Correction AI Tool: Harvard Medical School unveiled this tool after identifying bias in AI models used to analyse pathology samples. The new model reduces bias and improves cancer diagnosis across populations.
OncoMark: Scientists at the S. N. Bose National Centre for Basic Sciences, an autonomous institute under the Department of Science and Technology (DST), in collaboration with Ashoka University, unveiled this open-source AI model aimed at enabling personalised cancer therapy. Trained on a dataset of 3.1 million single cells across 14 cancer types, OncoMark predicts how hallmarks such as metastasis, immune evasion, and genomic instability interact to drive tumour growth and therapy resistance.
Smart Doctor: Developed by AIIMS New Delhi, Smart Doctor is an AI-powered clinical decision support system (CDSS) planned for deployment across 70,000 public and private hospitals nationwide. The system aims to reduce medical errors and enhance the quality of care by supporting and standardising diagnostic and treatment decisions, particularly for long-term and non-communicable diseases.
MadhuNETrAI: The All India Institute of Medical Sciences (AIIMS), in collaboration with the Union Health Ministry’s e-health division and Wadhwani AI, launched this tool to enable non-specialist health workers to conduct diabetic retinopathy screening using AI. The model went live in late 2025 across 38 centres in 11 states.
Garbhini-GA2: This India-specific model, developed to accurately determine fetal age during the second and third trimesters, was unveiled in 2024, but clinical validation and deployment efforts continued into 2025. While Western models often misestimate gestational age in Indian pregnancies by 7–10 days, Garbhini-GA2 reduces the error to under 0.5 days—crucial for predicting preterm births in India. It was developed by BRIC-THSTI, Faridabad, and IIT Madras.
