Large Language Models (LLMs) have emerged as powerful general-purpose reasoning systems, yet their development remains dominated by English-centric data, architectures, and optimization paradigms. This exclusionary design results in structural under-representation of linguistically diverse regions such as India, where over 20 official languages and 100+ dialects coexist alongside phenomena like code-switching and diglossia. We introduce PARAM-1, a 2.9B parameter decoder-only, text-only language model trained from scratch with an explicit architectural and linguistic focus on Indian diversity. PARAM-1 is trained on a bilingual dataset consisting of only Hindi and English, constructed with a strong focus on fact-rich, high-quality content. It is guided by three core principles: equitable representation of Indic languages through a 25% corpus allocation; tokenization fairness via a SentencePiece tokenizer adapted to Indian morphological structures; and culturally aligned evaluation benchmarks across IndicQA, code-mixed reasoning, and socio-linguistic robustness tasks. By embedding diversity at the pretraining level-rather than deferring it to post-hoc alignment-PARAM-1 offers a design-first blueprint for equitable foundation modeling. Our results demonstrate that it serves as both a competent general-purpose model and a robust baseline for India-centric applications.
@article{arxiv.2507.13390,
title = {PARAM-1 BharatGen 2.9B Model},
author = {Kundeshwar Pundalik and Piyush Sawarkar and Nihar Sahoo and Abhishek Shinde and Prateek Chanda and Vedant Goswami and Ajay Nagpal and Atul Singh and Viraj Thakur and Vijay Dewane and Aamod Thakur and Bhargav Patel and Smita Gautam and Bhagwan Panditi and Shyam Pawar and Madhav Kotcha and Suraj Racha and Saral Sureka and Pankaj Singh and Rishi Bal and Rohit Saluja and Ganesh Ramakrishnan},
journal= {arXiv preprint arXiv:2507.13390},
year = {2025}
}