The organisations behind the models

An open-source research lab at IIT Madras advancing AI for Indian languages. Its datasets, benchmarks, and models — like Airavata and IndicTrans — are widely used in research and industry.

An Indian AI company building voice-first, multilingual models made for India. Behind open releases like Sarvam-1, with a focus on practical, affordable AI.

A Bengaluru-based, open-source-first AI research lab founded by Aditya Kolavi. Built Ambari, one of the first bilingual Kannada-English models, plus open tools like OmniParse and the multilingual OCR project Nayana.

The AI division of Ola, building large multilingual models and a consumer AI assistant for Indian users, along with developer tools across many Indian languages.

An enterprise conversational-AI company founded in 2016, known for chatbot technology used across banking and government. BharatGPT is its generative-AI platform, adding multilingual voice and text support in a dozen-plus Indian languages.

A Government of India initiative (under MeitY) to make digital services available in every Indian language, sharing open language data and tools for translation and speech.

A small team building multilingual voice and chat assistants for Indian farmers, founded by computer scientist Pratik Desai. Its Dhenu models are fine-tuned specifically on agricultural knowledge and real farmer conversations.

A joint venture between Abu Dhabi-based investment firm 3AI Holding and SML India (Seetha Mahalaxmi Healthcare) behind Hanooman, a multilingual generative-AI platform aimed at healthcare, governance, and financial services.

A small, community-style AI team best known for Kan-LLaMA (Kannada Llama) — a Llama-2 model adapted for Kannada on a modest budget, showing how far open fine-tuning can go with limited resources.

A community group focused on open models for Marathi and nearby languages, bringing strong, free AI to Marathi speakers, local media, and small businesses.

Google's dedicated research lab in Bengaluru. Released MuRIL, an open multilingual model covering 17 Indian languages and their transliterated forms, used to improve Indian-language understanding in products like Google Assistant.

A community-led effort building open models for Telugu and other Indian languages, often by fine-tuning leading open models — like the Navarasa series — for strong local performance.

A volunteer research community, part of the wider Odisha.ML network, building open models and datasets for Odia — a language with comparatively few AI resources.

A collaboration between Sarvam AI, a fast-growing Indian startup building sovereign, voice-first Indic models, and AI4Bharat, the IIT Madras research lab its founders came from. Together they released OpenHathi, an early open Hindi LLM.

India's first government-supported multimodal LLM initiative, anchored by IIT Bombay with a consortium of leading IITs and IIITs under the National Mission on Interdisciplinary Cyber-Physical Systems.

The in-house AI initiative of Tech Mahindra, a large IT services company. Project Indus is its foundational LLM effort for Hindi and its many dialects, built with hardware partners including Dell and NVIDIA.

A multilingual-AI startup founded by Pranav Mistry, formerly president of Samsung's STAR Labs, backed by Naver and Reliance Jio. Sutra is its LLM family built for South and Southeast Asian languages, including several under-resourced Indian languages.

An independent developer behind Tamil-Llama, adapting popular open models for fluent Tamil — a great example of how individual builders strengthen the ecosystem.