Sarvam AI’s participation in the bid is set to significantly impact India’s AI landscape, especially after DeepSeek’s success in developing a large language model with limited resources, raising questions about the need for massive investment.

Bengaluru-headquartered generative AI startup Sarvam AI is preparing to submit a proposal to the Ministry of Electronics and Information Technology (MeitY) as part of the IndiaAI Mission. This initiative seeks to promote the development of AI foundational models tailored specifically for India, trained on local datasets, marking a significant step toward building an India indigenous AI model.

Pratyush Kumar, co-founder of Sarvam AI, stressed the importance of creating homegrown AI models for India’s future. Speaking at a recent business conclave, he said, “The rise of DeepSeek has made us question where our own models are. India needs to focus on developing AI models for India, or we risk being left behind in the global AI race, just as we have in previous technological generations.”

Also Read: India Partners with Sarvam AI to Develop Sovereign AI SOlutions

In response to the growing demand for local AI solutions, the IndiaAI Mission launched a proposal on January 31, inviting Indian startups, researchers, and entrepreneurs to participate in developing AI foundational models for India. The Ministry plans to provide equity-based funding and computational resources to the selected entities, ensuring strong support for companies like Sarvam AI in their efforts to contribute to the country’s AI sovereignty.

Leading India’s AI revolution

Sarvam AI, founded in 2023 by Vivek Raghavan and Pratyush Kumar, has already made significant advancements in AI, particularly with language models. The startup introduced Sarvam AI’s indigenous model, Sarvam 2B, India’s first open-source foundational small language model (SLM) focused on Indic languages. Trained on 4 trillion tokens, Sarvam 2B has been shown to outperform Meta’s Llama model in processing Indian languages, positioning Sarvam AI as a leader in the creation of AI models tailored for India.

In addition to Sarvam 2B, the company also launched Shuka 1.0, an open-source audio language model (audioLM) capable of mimicking and continuing audio input, such as musical beats or spoken languages. Shuka 1.0 was based on the Llama 8B model and was aimed at expanding audio processing capabilities within the Indian AI ecosystem.

Sarvam AI’s innovative work has been bolstered by a $41 million Series A funding round in December 2023, led by Lightspeed Venture Partners, with participation from Peak XV Partners and Khosla Ventures. However, creating AI foundational models remains a resource-intensive task. Despite this, Sarvam AI remains focused on producing indigenous AI models, a goal that aligns with India’s broader ambitions for AI sovereignty.

Sarvam AI’s involvement in the bid is expected to play a pivotal role in advancing India’s AI landscape. The success of DeepSeek, which developed a large language model with limited resources, has sparked industry debates over whether massive investment is always necessary. As Sarvam AI continues to push forward with its mission, its efforts to develop efficient, affordable, and localized AI models will be crucial for India’s future in the AI space.