India’s own AI: Building a multilingual, inclusive intelligence model

India is moving rapidly toward launching its first homegrown AI model in the next 6 to 8 months. This foundational model—being developed with Indian datasets and tailored for local languages and societal needs—aims to make AI truly accessible and impactful.

Backed by initiatives like BharatGen, BHASHINI, and Sarvam-1, and supported by a large-scale GPU infrastructure, India is focusing on creating Large Language Models (LLMs) and multimodal systems for public-good applications. With strong government support and open participation from startups and researchers, this AI mission is set to power sectors like healthcare, agriculture, and education—shaping a future where technology works for every Indian.

Image Source: Freepik

India is working ambitiously in developing its own foundational artificial intelligence (AI) models, tailored to address the country’s unique linguistic, cultural, and sectoral needs. The government is actively facilitating the creation of Large Language Models (LLMs) and domain-specific AI solutions that reflect the nation’s diversity. In support of this vision, multiple Centres of Excellence have been set up to foster advanced AI research and innovation.

A major milestone in this journey is the IndiaAI initiative, which has invited proposals for developing indigenous foundational models, including LLMs and Small Language Models (SLMs). These models are expected to play a crucial role in solving India-specific problems by leveraging local datasets and real-world applications.

One such initiative is Digital India BHASHINI, an AI-led language platform that aims to make the internet and digital services accessible in Indian languages. It enables voice-based access and content creation, promoting inclusivity across regions. Another key development is BharatGen, launched in 2024 in Delhi, which is the world’s first government-funded multimodal LLM initiative. BharatGen brings together a consortium of top AI researchers from Indian academic institutions and focuses on enhancing public service delivery and citizen engagement through foundational models in language, speech, and computer vision.

In parallel, Sarvam-1, a large language model optimised for Indian languages, has emerged as a critical tool for language translation, summarisation, and content generation. With 2 billion parameters and support for ten major Indian languages, Sarvam-1 reinforces the localisation aspect of India’s AI efforts. Open-source innovation is also thriving—Chitralekha, developed by AI4Bhārat, is a video transcreation platform that allows users to generate and edit audio transcripts in multiple Indic languages. Meanwhile, Hanooman’s Everest 1.0, a multilingual AI system developed by SML, currently supports 35 Indian languages, with plans to expand to 90.

India is now gearing up to launch its indigenously developed AI model within the next 6 to 8 months—an ambitious initiative aimed at delivering highly innovative and contextualised solutions tailored for the country’s diverse linguistic and cultural landscape. As part of this broader mission, the government is actively inviting participation from startups, researchers, and entrepreneurs to co-develop advanced AI models rooted in Indian data ecosystems. With a strong focus on collaboration and innovation, the effort is expected to result in cutting-edge AI systems designed to serve national priorities.

To support the development of indigenous AI models, the government has built a robust computing infrastructure equipped with 18,693 GPUs—including H100, H200, and MI 200 300 units—far exceeding the initial targets. A key pillar of this initiative is affordability. India’s compute facility is being offered at highly competitive rates, with GPU usage priced at just ₹115.85 per hour—significantly lower than the global average of US$ 2.5–$3 per hour. Furthermore, with a 40% government subsidy, the cost of high-end computing can be brought down to under ₹100 per hour, making advanced AI training accessible to a broader ecosystem of innovators.

The mission goes beyond technological advancement and places a strong emphasis on social impact. It aims to develop homegrown AI-powered solutions for critical sectors such as agriculture, healthcare, weather forecasting, and disaster management. So far, 18 real-world applications have been identified, targeting national challenges like climate change, learning disabilities, and agritech enhancement—underscoring the commitment to making AI a force for inclusive development and public welfare.

How does India compare globally?

While India is building its AI model with a strong focus on linguistic diversity, public benefit, and cost-effectiveness, other nations are also racing ahead with their own sovereign AI efforts.

The United States dominates with powerful models like OpenAI’s ChatGPT, Anthropic’s Claude, and Google’s Gemini, all backed by immense compute and capital. China, meanwhile, has launched DeepSeek and Ernie Bot, trained on extensive national datasets and aligned with state objectives.

France is investing in Mistral AI to reduce dependency on U.S. models, while Gulf nations are establishing their digital sovereignty through models like the UAE’s Falcon and Saudi Arabia’s ALLaM, backed by strategic investments in companies such as G42 and collaborations with Nvidia and AWS.

In comparison, India’s approach is uniquely democratic and inclusive—grounded in public-private partnerships, affordability, multilingual capabilities, and a focus on real-world applications. With an open call for participation and a massive compute backbone, India is not just catching up, but creating a model of responsible, citizen-first AI innovation.