Spotlight on India’s Thriving Open-Source AI Ecosystem: Top Models Competiting Globally

Overview of India’s AI Landscape

India’s open-source AI sector is rapidly evolving, showcasing a diverse array of homegrown models that excel in translation, speech recognition, and large language processing. Initiatives from IndiaAI and AIKosh have further accelerated this growth, establishing India as a formidable player on the global AI stage.

Comprehensive List of Leading Open-Source AI Models in India (As of October 2025)

India’s Booming Open-Source AI Ecosystem: A Comprehensive Overview

India’s open-source AI ecosystem has rapidly evolved, positioning itself as a formidable player on the global stage. With homegrown models in translation, speech recognition, and large language processing (LLP), India is witnessing a surge fueled by initiatives like IndiaAI and AIKosh. This growth is not merely anecdotal; it’s backed by impressive research across platforms such as Hugging Face, GitHub, and AIKosh.

As of October 2025, here’s a look at some of the most downloaded and widely used open-source AI models spearheading India’s AI revolution.

IndicTrans2 — AI4Bharat

India’s flagship translation model, IndicTrans2, stands out by supporting all 22 scheduled languages and over 110 translation directions. The En-Indic 1B variant enjoys approximately 13,554 downloads per month on Hugging Face, making it the go-to model for translation tasks involving Indian languages.

IndicBERT — AI4Bharat

IndicBERT is a multilingual ALBERT model trained on 12 Indian languages, boasting around 9 billion tokens. With close to 18,444 downloads per month, it is widely used for sentiment analysis, text classification, and entity recognition, showcasing its versatility in natural language processing (NLP) tasks.

Airavata — AI4Bharat

The Airavata model is tailored for Hindi instruction and is fine-tuned from OpenHathi. It performs admirably on Hindi NLP benchmarks, registering around 281 monthly downloads. This model plays a crucial role in enhancing Hindi language applications.

IndicWav2Vec — AI4Bharat

Representing the broadest linguistic diversity in Indian automated speech recognition (ASR), IndicWav2Vec is trained on 40 Indian languages. The Hindi model alone garners about 1,997 monthly downloads, making it a significant player in speech recognition.

Sarvam-1 — Sarvam AI

Sarvam-1 is a two-billion-parameter language model optimized for 10 major Indic languages, including Hindi, Tamil, Bengali, and Marathi. Released by Sarvam AI, the model marks significant strides in multilingual AI applications across the Indian landscape.

Sarvam-M — Sarvam AI

Coming in at 24 billion parameters, Sarvam-M is designed for reasoning tasks in Indic languages. Despite facing controversies regarding download manipulation, it continues to be a noteworthy model in India’s AI scene.

OpenHathi-7B — Sarvam AI

One of India’s original large language models (LLMs), OpenHathi-7B is a Hindi-English bilingual model based on Llama architecture. With an average of 1,733 monthly downloads, it competes strongly with models like GPT-3.5 for Hindi language tasks.

Krutrim-1 — Krutrim AI

Krutrim-1 employs Mistral architecture to deliver a seven-billion-parameter multilingual foundation model, trained on a staggering two trillion tokens across Indic languages. It forms a core part of Krutrim’s growing AI offerings.

BharatGPT-3B-Indic — CoRover.ai

As a transformer-based language model optimized for Indic languages, BharatGPT-3B-Indic crossed 2,000 downloads within days of its release. Its quantized GPT-generated versions receive around 129 monthly downloads, illustrating its rapid adoption.

Param-1 — BharatGen AI

Param-1 marks a milestone as one of the first Indian LLMs built from the ground up by the BharatGen consortium. This 2.9-billion-parameter bilingual foundation model has seen about 4,369 monthly downloads, ranking it among the most utilized Indian LLMs.

Project Indus — Tech Mahindra

Project Indus is an open-source Hindi LLM supporting 37 dialects and focused on enterprise applications and local data comprehension. This model has already seen the release of a second version, further enhancing its capabilities.

MuRIL — Google Research India

The MuRIL model, a BERT-based model trained on 17 Indian languages, remains one of the most widely adopted Indian language NLP models on a global scale. Its effective results have made it a go-to resource for many developers.

Vakyansh ASR — EkStep Foundation

Vakyansh ASR is an automatic speech recognition model trained for multiple Indian languages. The Hindi version gets around 2,838 monthly downloads, while others cover languages like Tamil, Kannada, Telugu, and Bengali.

IndicTTS — C-DAC/IIT Madras Consortium

The IndicTTS model is an open-source text-to-speech solution supporting 13 languages, including Hindi, Tamil, Malayalam, and Marathi. Its widespread adoption in digital voice applications underscores its significance.

Indic Parler-TTS — AI4Bharat

Leading the way in Indian text-to-speech downloads with about 12,295 downloads per month, Indic Parler-TTS is a multilingual text-to-speech model that supports 21 languages and has been trained on extensive data from unique voices.

Paramanu — Gyan AI

The Paramanu family consists of lightweight models optimized for minimal computing needs across 10 Indian languages. Notably, it includes Paramanu-Ayn, India’s first legal AI model tailored for legal texts, and Paramanu-Ganita, designed for Indian educational applications.

Chitrarth — Krutrim AI

Chitrarth serves as a vision-language model combining the capabilities of Krutrim-1 and a SigLIP visual encoder, supporting 10 Indic languages alongside English for effective multimodal reasoning.

e-vikrAI — BharatGen

e-vikrAI focuses on Indian e-commerce, automating product cataloging through image and text inputs while ensuring cultural accuracy. This innovation streamlines the cataloging process for sellers, reducing manual labor significantly.

The impressive list of models illustrates how India’s open-source AI ecosystem is not just growing, but thriving. The collaboration of startups, academia, and government initiatives is setting the stage for innovative applications that will ultimately benefit millions, while also contributing to the global AI community. With a focus on multilingual capabilities and local contexts, India’s AI future looks promising and poised for further advancement.