Sarvam AI To Bid For Building Indigenous AI Foundational Model

Bengaluru-based GenAI startup Sarvam AI plans to submit a proposal to the electronics and IT ministry (MeitY) to build an indigenous AI foundational model under the IndiaAI Mission.

Speaking at NDTV Profit Conclave, Sarvam AI cofounder Pratyush Kumar said, “We (Sarvam AI) will participate in the call for proposals put out by the ministry.” 

The publication reported that Kumar highlighted the importance of having Indian AI models. “DeepSeek has made us question ‘where are our models’. We need to think about India’s sovereign AI models, else we have the risk of being left out if we don’t step up, as has happened in the previous generations,” he was quoted as saying. 

On January 31, the IndiaAI Mission floated a proposal seeking bids from Indian startups, researchers and entrepreneurs to collaborate on building foundational models trained on Indian datasets. 

Under the proposal, the Centre is mulling offering equity-based funding as well as compute credits for the selected entity(s) that will undertake this task. 

Notably, Sarvam AI was founded in 2023 by Vivek Raghavan and Pratyush Kumar. Since then, the startup has launched small language models (SLMs). 

Earlier, the startup claimed to have launched India’s first open source foundational SLM, Sarvam 2B, focussed on Indic languages. The startup claims that it is trained with 4 Tn tokens and performs better than Meta’s Llama in Indian language. 

Prior to that, Sarvam AI launched Shuka 1.0, which it said was India’s first open source audioLM, which can mimic or continue the audio input given – be it a musical beat or a particular language. However, it was an extension on the Llama 8B model. 

The startup raised $41 Mn (around INR 342 Cr) in its Series A funding round led by Lightspeed Venture Partners, in participation with Peak XV Partners and Khosla Ventures, in December 2023

This was among one of the largest fundraisers by Indian AI startups. However, building AI foundational models requires a lot of resources, which might be a constraint for Sarvam AI.

Earlier, Sarvam cofounder Raghavan told Inc42 that the startup is “reasonably frugal” in its approach but GenAI is a cost-intensive business. 

However, the emergence of DeepSeek has raised questions if companies need to spend big amounts to develop LLMs. Last month, DeepSeek sent shockwaves in the technology ecosystem when it managed to develop its own LLM with only 2,000 NVIDIA H800 GPUs.

The post Sarvam AI To Bid For Building Indigenous AI Foundational Model appeared first on Inc42 Media.