BharatGen will unveil Param2 17B at the India AI Impact Summit 2026, a 17 billion parameter multilingual AI model engineered for 22 Indian scheduled languages. Constructed using a Mixture of Experts (MoE) architecture and trained extensively on India-specific datasets, the model excels in reasoning, mathematics, coding, text generation, speech processing, and vision tasks. Supported by the IndiaAI Mission, BharatGen accesses government-backed high-performance computing infrastructure and the Bharat Data Sagar repository, enabling open deployment for developers, startups, researchers, and enterprises in governance, healthcare, education, finance, and public services.
Government Pilots Target Citizen Services and Governance
Live demonstrations at the summit showcase practical integrations. MahaGPT, developed with the Government of Maharashtra and Maharashtra Institution for Transformation (MITRA), streamlines complex urban development and revenue administration processes through AI-driven efficiency and accuracy. The central government’s Department of Water and Sanitation employs multimodal tools for citizen access to services. Goa Electronics Limited advances AI-led digital transformation initiatives for the Government of Goa, focusing on scalable public sector applications.
BharatGen CEO Rishi Bal stated that these models prioritise real-world deployment across citizen services while bolstering India’s sovereign AI capabilities. The summit facilitates fast go-to-market pathways alongside open access for ecosystem builders.
Healthcare, Education, and Cultural Preservation Applications
In healthcare, Medsum by Mata Amrita Technologies leverages BharatGen models to bridge doctors and patients via structured health records and natural conversational interfaces. Kotak Education Foundation deploys a fluency evaluation tool assessing spoken English skills, generating detailed performance metrics and learner insights to enhance communication training.
Cultural heritage benefits from Gyan Bharatam, a Ministry of Culture digitisation programme using BharatGen’s OCR and conversational AI to transcribe and process ancient manuscripts across 22 languages. The initiative aims for a national archive encompassing 50 crore documents, supporting preservation, translation, and scholarly access through a central Delhi headquarters and regional centres.
12 Foundational Models Spotlight India’s AI Ecosystem
The summit features announcements from 11 additional organisations unveiling foundational models. Sarvam AI’s recent Saaras V3 advances speech recognition for code-mixed and noisy environments across all 22 scheduled Indian languages plus English. Powered by a new architecture with native streaming, configurable decoding modes (Accurate, Balanced, Fast), automatic language detection, numeral formatting control, and speaker diarization, Saaras V3 achieves 19 percent word error rate on IndicVoices benchmarks.
Param2 17B advances India’s Viksit Bharat 2047 vision for technological self-reliance, embedding Indic data sovereignty into large-scale AI foundations.
