BharatGen

What BharatGen does

BharatGen Technology Foundation is India's first government-backed sovereign generative AI ecosystem, established in October 2024 as a consortium project led by IIT Bombay's Technology Innovation Hub (TIH) under the National Mission on Interdisciplinary Cyber-Physical Systems (NM-ICPS), Department of Science and Technology (DST), and incorporated as a Section 8 non-profit entity in November 2025 (CIN U74909MH2025NPL460506). The organization builds a family of multilingual and multimodal foundation models — led by the Param-2 17B-parameter Mixture-of-Experts model (activating 2.4B parameters per token, trained on ~22 trillion tokens covering English and all 22 Scheduled Indian languages), the Param-1 family (2.9B/7B bilingual Hindi-English), four domain-tuned variants (AyurParam, AgriParam, FinanceParam, LegalParam), Patram 7B for vision/document understanding, Shrutam 2 for automatic speech recognition across 12 Indian languages with code-mixed support, and Sooktam 2 for reference-guided text-to-speech. The technology stack rests on a custom morphology-aware tokenizer that reduces token-to-word ratios by ~6% across 17 Indic languages, and is trained on Bharat Data Sagar — positioned as India's largest India-centric dataset repository targeting 15,000+ hours of annotated voice data across 22 Indian languages by Q4 2025.

BharatGen serves Indian central and state government, defence, healthcare, agriculture, education, media, and enterprise customers via enterprise field sales, system-integration partnerships (IBM Watsonx/Red Hat OpenShift AI, L&T Semiconductor Technologies and L&T-Vyoma, NVIDIA), and consortium academic distribution. Named engagements include the Government of Andhra Pradesh (Telugu-first Swadeshi AI Tech Hub), Government of Maharashtra (MahaGPT over 150,000+ Government Resolutions), the Indian Defence Ministry (active integration discussions), Amrita Hospitals (Med-Sum clinical AI Scribe deployed at Faridabad and Kochi), India Today Group (Sutra AI news anchor), and Kotak Education Foundation.

The business model is grant-funded rather than commercially monetized: BharatGen received ₹235 crore from NM-ICPS/DST in 2024 and ₹988.6 crore (the largest single allocation) from MeitY under the IndiaAI Mission in September 2025, structured as a 40% grant / 60% recoverable arrangement. Models, datasets, and benchmarks are released open-source on Hugging Face and AI Kosh, with selective licensing for advanced assets to government and trusted partners. Revenue is therefore characterized as Pre-revenue in a commercial sense, with monetizable sovereign-AI contracts (defence, healthcare, enterprise) implied as the future transition path beyond grant funding.

Name

BharatGen

Legal name

BharatGen Technology Foundation

Website

https://bharatgen.com

Company type

Private

Founded year

2024

Operating status

Operating

Headcount range

51–100 employees

Short description

BharatGen is India's first government-backed sovereign AI non-profit, building multilingual and multimodal foundation models (Param-2 17B MoE, Param-1, Patram, Shrutam, Sooktam) covering 22 Indian languages for central and state government, defence, healthcare, and enterprise customers.

Ownership category

akta.pro rank

Product category

Foundation Models / Sovereign Generative AI Platform

NAICS

Software Publishers (5132), Scientific Research and Development Services (5417), Computer Systems Design and Related Services (5415)

SIC

Services-Prepackaged Software (7372), Services-Computer Programming, Data Processing, Etc. (7370)

akta.pro primary industry

Multimodal Generation (Text-Image-Video-Audio) & Creative Toolchains (HDAAACAG)

akta.pro secondary industries

Enterprise Foundation Model Integration & APIs (Connectors, Governance, Deployment) (HDAAACAO), Speech Translation & Multilingual Speech Tech (HDAAAFAE), Machine Translation & Multilingual NLP (HDAAADAD), Open-Source Model Ecosystems & Model Marketplaces (HDAAACAM), Synthetic Data & Data Augmentation for Foundation Models (HDAAACAK), Fine-Tuning, Adaptation & Custom Model Training (PEFT/LoRA/RLHF) (HDAAACAC)

Keywords

• Multilingual foundation models
• Sovereign AI platform
• Automatic speech recognition
• Document understanding AI
• Indic language models

Headquarters

HQ city

Mumbai

HQ country

India

HQ region

Asia

Offices2 records

Markets served

GTM type

B2B

Offering type

Software

Cost components

Technology or R&D, Personnel, Infrastructure, Operations, Marketing or Sales, Supply Chain

Revenue model

1. Government Grant Funding (IndiaAI Mission / DST / NM-ICPS): Primary funding mechanism: ₹988.6 crore awarded by MeitY under the IndiaAI Mission, supplemented by ₹235 crore from NM-ICPS / Department of Science and Technology; structured as a 40% grant / 60% recoverable (equity/debt) arrangement per IndiaAI Mission rules, with funding dispersed across a ~1 year horizon to build models up to 1 trillion parameters.
2. Open-Source Model & Dataset Licensing (free tier): Models (Param-1, Param-2, Patram, Shrutam, Sooktam, domain Params) and datasets (Bharat Data Sagar, BhashaBench series, IndicParam, ParamBench) released openly on Hugging Face and AI Kosh for public use, research, and downstream application development.
3. Selective Licensing / Partnership Access to Advanced Resources: Advanced assets and certain models are reserved for selective access by government agencies and trusted partners under structured partnership arrangements (e.g., IBM enterprise-grade integration, government-of-India deployments).
4. Government & Enterprise Pilot / Co-Development Engagements: Pilot and integration revenues implied through co-development with state governments (Andhra Pradesh Telugu-first AI Tech Hub, Maharashtra MahaGPT), system integrators, and enterprise partners (IBM Watsonx/Red Hat OpenShift AI deployment).

Pricing tiers

Go-to-market motion1 record

Distribution channels7 records

Marketing channels10 records

Core offering

BharatGen develops and open-sources a family of multilingual and multimodal foundation models for Indian languages, including text LLMs (Param-1 bilingual and Param-2 17B MoE covering 22 Scheduled Indian languages plus English), domain-tuned variants (AyurParam, AgriParam, FinanceParam, LegalParam), a document vision-language model (Patram 7B), multilingual speech recognition (Shrutam 2), and multilingual text-to-speech (Sooktam 2). It also builds the Bharat Data Sagar India-centric dataset and deploys sovereign AI infrastructure via partnerships with IBM, L&T, and NVIDIA for government and enterprise customers.

Product overview

BharatGen is India's first government-backed sovereign generative AI ecosystem, anchored at the Technology Innovation Hub (TIH) at IIT Bombay under the Department of Science and Technology's National Mission on Interdisciplinary Cyber-Physical Systems (NM-ICPS) and funded under the IndiaAI Mission with ₹988.6 crore. Its platform-plus-modules architecture spans multilingual foundation models, domain-specific variants, and sector applications. The text-model stack is led by Param2 (17B Mixture-of-Experts, ~22T tokens, 22 Scheduled Indian languages plus English) with the Param1 family (2.9B and 7B bilingual Hindi-English models), and four domain-tuned derivatives AyurParam, AgriParam, FinanceParam and LegalParam. The vision stack centers on Patram 7B for document understanding. The speech stack comprises Shrutam 2 (multilingual ASR) and Sooktam 2 (multilingual TTS), with Shrutam 1 Hindi for Hindi ASR. Bharat Data Sagar is the underlying India-centric dataset initiative, and applied products include Krishi Sathi (agriculture farm bot), e-VikrAI (e-commerce seller assistant), DocBodh (document understanding), Sutra (AI news anchor with India Today Group), MahaGPT (Maharashtra governance Q&A), Med-Sum (healthcare with Amrita Institutions), and Bharat-VISTAAR (farmer advisory).

Differentiator

Problem solved

Functional benefit

Brands

• Param: BharatGen's family of multilingual text foundation models (Param-1 and Param-2) for Indian languages.
• Shrutam
• Sooktam
• Patram
• DocBodh
• Bharat Data Sagar
• Krishi Sathi
• e-VikrAI

Products and services

• Param-2 (17B Mixture-of-Experts) 17-billion-parameter Mixture-of-Experts multilingual foundation model that activates 2.4B parameters per token, trained on approximately 22 trillion tokens covering English and all 22 Scheduled Indian languages, with reasoning, chain-of-thought, tool calling, mathematics, and code generation capabilities. Open-source on Hugging Face.
• Param-1 (2.9B Bilingual Hindi-English) Bilingual 2.9-billion-parameter Hindi-English foundation model pretrained from scratch on 7.5 trillion tokens (2.77T Hindi) with a custom 128K-vocabulary SentencePiece tokenizer optimized for Indian scripts, including an instruction-tuned checkpoint for governance, education, healthcare, legal, and agriculture use cases. Open-source on Hugging Face and AI Kosh.
• Param-1 7B 7-billion-parameter scale-up of the Param-1 bilingual Hindi-English foundation model, part of BharatGen's foundational text model family released on Hugging Face.
• AyurParam Domain-tuned Param model specialized for Ayurvedic and classical medical texts, supporting sector-specific reasoning on Indian-context Ayurveda queries. Standalone downloadable model on Hugging Face.
• AgriParam Domain-tuned Param model specialized for Indian agricultural knowledge and farmer-facing use cases. Standalone downloadable model on Hugging Face.
• FinanceParam Domain-tuned Param model specialized for Indian financial queries and government financial examination domains. Standalone downloadable model on Hugging Face.
• LegalParam Domain-tuned Param model specialized for India-centric law across 14 legal domains. Standalone downloadable model on Hugging Face.
• Patram 7B (Document Vision Model) India's first document foundation model, combining a Vision Transformer with a 7B-parameter language model for multilingual image and document understanding, visual question answering, table recognition, and document analysis in Indian scripts. Released on Hugging Face and AI Kosh.
• Shrutam 2 (Multilingual ASR) LLM-powered automatic speech recognition model that transcribes speech across 12 Indian languages (Hindi, Marathi, Tamil, Telugu, Malayalam, Kannada, Odia, Bengali, Urdu, Assamese, Gujarati, Punjabi) with support for code-mixed Hindi-English audio. Open-source on Hugging Face and AI Kosh.
• Sooktam 2 (Multilingual TTS) Reference-guided text-to-speech model that synthesizes natural expressive speech across 12 Indian languages (Hindi, Marathi, Gujarati, Tamil, Telugu, Kannada, Bengali, Malayalam, Odia, Urdu, Punjabi) plus Indian English while preserving speaker voice characteristics. Open-source on Hugging Face and AI Kosh.
• Shrutam 1 Hindi (Hindi ASR) Hindi-language automatic speech recognition model from the Shrutam 1 generation, released on Hugging Face and AI Kosh for Hindi speech transcription.
• Bharat Data Sagar BharatGen's flagship India-centric data initiative building the world's largest dataset of Indian text, speech, and images covering Indian languages, culture, history, and philosophy, with 15,000+ hours of annotated voice data across 22 Indian languages targeted by Q4 2025. Used to train BharatGen's text, ASR, TTS, and other foundation models.
• Krishi Sathi India's first farm bot for Indian farmers, equipped with text-to-speech capability and data-driven insights to guide and support farmers in agriculture. Featured on BharatGen's products page.
• e-VikrAI AI assistant for Indian sellers that supports product cataloguing and business operations across India's e-commerce landscape. Featured on BharatGen's products page.
• DocBodh Document understanding product built on BharatGen's Patram vision-language model for analysing, extracting and reasoning over Indian-language documents. Featured on BharatGen's products page.
• Sutra (AI News Anchor) AI-driven news anchor developed by India Today Group in collaboration with BharatGen, designed to synthesize complex real-time information and produce sovereign, contextual summaries with culturally aware and linguistically inclusive AI capabilities for Indian audiences. Launched at India AI Impact Summit 2026.
• MahaGPT Compact language model launched by the Maharashtra Institution for Transformation (MITRA), IIT Bombay and BharatGen that queries over 150,000 Maharashtra Government Resolutions and regulations on land and urban development to strengthen evidence-based decision-making and policy transparency. Launched at the AI Impact Summit 2026.
• Med-Sum (Medical Summary via AI Scribe) Healthcare AI solution co-developed with Amrita Institutions providing real-time clinical transcription and multilingual patient-friendly explanations. Already in clinical use at Amrita Hospitals in Faridabad and Kochi; part of the broader BharatGen-Amrita Medical Foundation Models and EMR framework collaboration.

Quantifiable outcome

• Largest beneficiary of IndiaAI Mission — ₹988.6 crore funding allocation (highest among 12 selected startups)
• +7 more outcomes

Named customers7 records

Segments6 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

Integration3 records

AI capability10 records

Feature10 records

Partnerships

17 partnerships are on record, tiered core, flagship and minor.

• NVIDIAcoreTechnology or Integration · 21 May 2026BharatGen builds on NVIDIA's Nemotron toolkit for sovereign AI applications spanning finance, healthcare, and industry. NVIDIA supports BharatGen with NeMo Curator, CUDA-X, Omniverse, and India-specific datasets like Nemotron-Personas-India (21M personas) and powers BharatGen's 17B-parameter Param-2 model. NVIDIA also backed BharatGen with over 20,000 Blackwell Ultra GPUs deployment via Yotta/L&T/E2E.
• Indian Defence MinistryflagshipStrategic or Co-development Partner · 11 May 2026Active discussions with the Indian defence ministry (alongside SarvamAI) to integrate BharatGen's locally-made AI systems into India's military capabilities — aiming to create an Indian equivalent of Palantir, driven by strategic concerns over foreign AI dependency in defence.
• Government of MaharashtracoreStrategic or Co-development Partner · 6 March 2026MoU signed between Maharashtra government, IIT Bombay and BharatGen (alongside OpenAI and Sarvam AI Letters of Intent) to strengthen the state's AI and emerging technology ecosystem — building AI capacity among civil servants, integrating AI tools into administrative workflows, and transitioning skilling strategies to applied, skill-based domains.
• L&T Semiconductor Technologies (LTSCT)flagshipStrategic or Co-development Partner · 3 March 2026Five-year MoU with BharatGen to design custom AI ASIC and xPU silicon platforms optimized for BharatGen's multilingual LLMs and domain-specific models. Part of the sovereign AI compute platform partnership signed in the presence of the Principal Scientific Adviser to the Government of India.
• L&T-Vyoma (Larsen & Toubro data center business)flagshipStrategic or Co-development Partner · 3 March 2026Five-year MoU with BharatGen to provide hyperscale AI-ready data center infrastructure including a 30 MW facility in Kanchipuram, Tamil Nadu, to host BharatGen's foundational models on sovereign infrastructure. Three-year phased implementation roadmap with joint working groups.
• Government of Andhra PradeshflagshipStrategic or Co-development Partner · 21 February 2026MoU signed in the presence of Hon'ble Chief Minister Nara Chandrababu Naidu, Rishi Bal (CEO, BharatGen) and Prof. Ganesh Ramakrishnan (Lead PI) to establish a state-wide Telugu-first AI Tech Hub. Initiative aligns with DST, MeitY, and IndiaAI to deliver citizen-centric AI for farmers, fishermen, entrepreneurs, administrators and students.
• Maharashtra Institution for Transformation (MITRA)coreStrategic or Co-development Partner · 21 February 2026Collaboration with IIT Bombay and BharatGen at AI Impact Summit 2026 to launch MahaGPT — a compact language model that queries 150,000+ Maharashtra Government Resolutions (GRs) and land/urban development regulations to enable evidence-based decision-making and citizen participation.
• NxtGen Cloud TechnologiescoreTechnology or Integration · 20 February 2026Partner in the Andhra Pradesh Swadeshi AI Stack (alongside IBM and BharatGen) — provides sovereign cloud and GPU infrastructure for the state-wide Telugu-first AI Tech Hub, enabling Telugu multilingual AI models deployed on sovereign infrastructure.
• AI Alliance (Project Tapestry)flagshipStrategic or Co-development Partner · 19 February 2026BharatGen anchors India's role in Project Tapestry, a global open consortium led by the AI Alliance and launched at the G7 Summit in Évian, France (June 19, 2026). BharatGen co-leads workstreams focused on distributed model training alongside participating nations including the US, France, India, Vietnam, Japan, Switzerland, and UAE.
• Amrita Institutions (Amrita Hospitals)coreStrategic or Co-development Partner · 19 February 2026MoU to co-develop sovereign, multilingual AI solutions for India's healthcare sector — Medical Foundation Models, EMR frameworks, and speech-first multimodal systems. First milestone is Med-Sum (Medical Summary via AI Scribe), already used by hundreds of doctors and patients at Amrita Hospitals in Faridabad and Kochi following clinical validation.
• India Today GroupcoreGTM or Marketing Partner · 17 February 2026Collaboration to develop Sutra, an AI-driven news anchor debuted at India AI Impact Summit 2026. Uses BharatGen's sovereign, culturally aware and linguistically inclusive models with an "AI Sandwich" architecture (human editorial intent at start, verification at end) to make high-level policy discussions accessible to the public.
• Kotak Education Foundation (KEF)minorStrategic or Co-development Partner · 16 February 2026MoU to collaborate on advancing AI in India's education sector — building and deploying indigenous, multilingual AI solutions covering AI-enabled assessment, multilingual content delivery, hyper-personalized learning tools, and data-driven insights; KEF handles outreach and implementation across Indian classrooms.
• Shashi Shekhar Vempati (Independent Director, BharatGen Board)minorOthers · 26 January 2026Former Prasar Bharati CEO and IIT Bombay alumnus Shashi Shekhar Vempati appointed as Independent Director on the board of BharatGen Technology Foundation, recognizing his career spanning public broadcasting, education governance, and technology advocacy.
• IBMflagshipStrategic or Co-development Partner · 17 September 2025Strategic collaboration to advance AI adoption in India powered by BharatGen's sovereign multimodal and LLMs. Joint work includes developing Indic solution templates (RAG, targeted domains) on IBM Watsonx and Red Hat OpenShift AI, building scalable Indic data pipelines, governance frameworks, Indic benchmarks, and new model architectures. Partners include IBM, Zoho, and Nasscom per CEO commentary.
• IIT Bombay (consortium lead)flagshipStrategic or Co-development PartnerBharatGen Technology Foundation is anchored under the Technology Innovation Hub (TIH) at IIT Bombay with Prof. Ganesh Ramakrishnan as Principal Investigator; IIT Bombay hosts BharatGen as a national sovereign AI ecosystem. BharatGen's registered office is at IIT Bombay, Powai, Mumbai.
• Consortium academic partners (IIT Madras, IIIT Hyderabad, IIT Kanpur, IIT Hyderabad, IIT Mandi, IIIT Delhi, IIT Kharagpur, IIM Indore)coreStrategic or Co-development PartnerEight additional consortium institutions contributing faculty and research to BharatGen's foundational model development, datasets, and benchmarks. Together with IIT Bombay, they form the academic backbone of India's sovereign AI ecosystem under NM-ICPS and the IndiaAI Mission.
• Yotta, L&T, E2E Networks, Netweb Technologies (cloud/infra partners)coreChannel Partner/ Reseller/ DistributorUnder the IndiaAI Mission's $1 billion+ initiative, Yotta (with L&T and E2E) deploys 20,000+ NVIDIA Blackwell Ultra GPUs to support BharatGen, Sarvam, Chariot and other Indian AI model builders. Netweb Technologies manufactures AI supercomputing systems domestically under Make in India.

Scale indicators13 records

Recent moves8 records

Expansion highlights7 records

Direct peers

• CoRover.ai: CoRover.ai is an India-based conversational AI platform offering multilingual chatbots and BharatGPT-style sovereign LLM products to Indian enterprises and government. It overlaps with BharatGen's enterprise/government go-to-market and Indic-language chatbot capabilities.
• Krutrim (Ola): Krutrim is Ola's India-focused AI venture building multilingual LLMs and conversational AI for Indic languages. It targets similar central and state government, enterprise and consumer use cases across India's 22 scheduled languages, overlapping directly with BharatGen's Param and Sooktam model lines.
• Sarvam AI: Sarvam AI is an India-based sovereign LLM startup and fellow IndiaAI Mission beneficiary building Indic-language foundation models. It directly competes with BharatGen for sovereign AI funding, state-government MoUs (including reported Indian Defence Ministry engagements) and Indic language benchmarks.
• Reverie Language Technologies: Reverie provides Indian-language localization, speech, and translation APIs powering Indic-language digital experiences. It is comparable to BharatGen on Indic speech (Shrutam/Sooktam) and Indic NLP capabilities, though focused more on localization than foundation model research.
• AI4Bharat (IIT Madras): AI4Bharat is an IIT Madras-hosted research initiative building open-source Indic language AI datasets, models and benchmarks (including IndicBERT, IndicTrans). It is the closest academic peer to BharatGen's consortium model and shares Bharat Data Sagar's mandate of open, India-centric language technology.

Broad incumbents

• Mistral AI: Mistral AI is a leading European open-weights foundation model lab whose efficient open-source models are a credible alternative to BharatGen's MoE-based Param-2 for cost-sensitive sovereign deployments. Comparable as an open-weight foundation model provider targeting enterprise and government use cases.
• IBM (Granite / watsonx): IBM is both BharatGen's flagship enterprise deployment partner (Watsonx, Red Hat OpenShift AI, Granite) and a broad incumbent in enterprise foundation models and AI services. Comparable as a strategic enterprise AI platform, though IBM's models are general-purpose and not India-specific.
• Meta AI (LLaMA): Meta's open-source LLaMA family is BharatGen's principal open-source foundation model benchmark and the implicit alternative for any developer evaluating open LLMs. Comparable as an open-weights LLM platform, though LLaMA allocates under 0.01% of training data to Indic languages versus Param-1's 25% Hindi allocation.

Emerging players

• Gnani.ai: Gnani.ai builds speech AI and multilingual voicebots for Indic languages targeting Indian banking, insurance and BFSI customers. It directly parallels BharatGen's Shrutam ASR and Sooktam TTS stack and its enterprise/BFSI go-to-market.
• Yellow.ai: Yellow.ai is an India-born enterprise conversational AI platform with multilingual support deployed across Indian and global enterprises. It overlaps with BharatGen on enterprise-grade multilingual chatbots and the sovereign-AI-for-enterprise narrative.

Market position

Strengths5 records

Weaknesses5 records

Competitive moat6 records

Key risks6 records

Key highlights7 records

Customer concentration

• Xx.com/BharatGen_Com
• YouTubeyoutube.com/@bharatgenofficial
• Facebookfacebook.com/bharatgen
• LinkedInlinkedin.com/company/bharatgen
• Instagraminstagram.com/bharatgen_com

Revenue estimate

Valuation estimate

Number of profiles

Profiles3 records

Funding overview

Funding rounds1 record

Investors1 record

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about BharatGen