Sarvam Edge: India is Getting Closer in its Quest for Sovereign AI
- Anupama Vijayakumar
- 2 hours ago
- 4 min read
A Made in India AI model unveiled at the AI Impact Summit has outperformed the leading players in the market in its ability to synthesize multiple Indian languages. Optimized to work in consumer devices and in low internet connectivity conditions, the Sarvam Edge epitomizes how linguistic and digital inclusivity can be embedded into design parameters. Its success in terms of improving accessibility goes to show how India can rewrite definitions for human-centric and inclusive AI.
India has finally arrived at the global Gen AI stage with Sarvam Edge, a homegrown Large Language Model (LLM) which can synthesize over 10 Indian languages with performance comparable to Google’s Gemini 3 and Open AI’s Chat GPT. Announced a few days before the AI Impact Summit, Sarvam Edge has been developed by Bengaluru-based Sarvam AI, a company which India’s Ministry of Electronics and IT selected in 2025 to build an indigenous AI model. Now, with Sarvam Edge gaining global acclaim at the AI Impact Summit, this marks a major milestone in India’s bid to build “Sovereign AI”, comparable to DeepSeek moments which in the recent past have upended the global AI ecosystem.
DeepSeek Moments and the Drive for Sovereign AI
The US-dominated AI marketplace has been shocked and awed by periodic “DeepSeek Moments” in the recent past. The phrase first captured popular imagination in January 2025 when Chinese company DeepSeek launched the DeepSeek R1 model. With performance parameters comparable to the leaders in the market including Chat GPT, DeepSeek R1 became highly popular, as evidenced by rising number of downloads from app stores. Since then, the phrase has come to represent an emerging player in the AI market thwarting the first movers’ advantage by building AI models with highly advanced capabilities. Such capabilities are achieved often using lesser resources as compared to the first movers, and at a fraction of the cost. DeepSeek, along with a handful of other Chinese companies have continued to deliver multiple similar fetes despite strict US export controls on chips.
In the realm of geopolitics, a DeepSeek moment can be seen as a modern-day equivalent of the Cold War era terminology on the “Sputnik Moment”. It essentially refers to the awe and sense of doom that fell upon American psyche when the Soviet Union became the first country to launch an artificial satellite in 1957. To the American intelligentsia, the Sputnik moment triggered a sense of unease, that Soviet Union had attained a first-mover advantage in spacefaring. It secondly cemented a sense of the need for the United States to catch up to the formidable Soviet prowess in technology.
The scenario today with US first-mover advantage in AI and compute is similar. Driven by national security concerns, a number of rising powers are striving to attain Sovereign AI. Competition in the military domain as well as threats to economic security including supply chain disruptions and market volatility is prompting countries to “produce artificial intelligence using its own infrastructure, data, workforce and business networks”.
Moreover, the pursuit of Sovereign AI, particularly in non-English speaking countries is also driven by a need to prevent linguistic exclusion. Addressing bias as well as making AI accessible has been a core driver of sovereign AI initiatives in in regions ranging from Asia to Africa and Latin America. Now with Sarvam Edge, India has marked a unique transition from evolving as a consumer of AI to being a creator.
Sarvam Edge: The India Way to Sovereign, Inclusive AI
Several unique traits of Sarvam Edge set it apart from its peers which are currently popular. Unlike Open AI or Google which operate on remote servers, Sarvam Edge runs entirely on consumer devices and hence, is not dependent on internet connectivity. This feature makes AI highly accessible, even from locations with sparse connectivity. Sarvam Edge is also equipped with advanced speech recognition capabilities for Indian languages including Hindi, Kannada, Telugu and Gujarati. It can process speech input in these languages and generate voice output. It can thereby facilitate cross-language communication. Particularly, in several benchmarks, its ability to synthesize Indian languages exceeds that of Google Speech-to-Text.
Moreover, inclusivity as a principle of AI ethics has been embedded into Sarvam Edge. The unique requirements of the Indian market in terms of accounting for its linguistic diversity have been largely ignored by leading players in generative AI, despite the country housing a large user base. Adapting these models trained on non-Indian data, hence subject Indian users to bias. Capabilities attained by homegrown models such as Sarvam AI are crucial for India’s quest to build AI for all. Moreover, at zero-cost per query, economic inclusivity is built into Sarvam AI. India envisages similar endeavours in Sovereign AI as important means of solving its challenges with respect to healthcare, education, agriculture and financial inclusion. Moreover, the advent of similar models in the next few years can help overcome linguistic barriers, bring about more cross-country integration and build societal resilience across the spectrum.
Disclaimer: The article expresses the author’s views on the matter and do not reflect the opinions and beliefs of any institution they belong to or of Trivium Think Tank and the StraTechos website.
Anupama is the Director (Research) of Trivium Think Tank, Thiruvananthapuram, Kerala, India. She is also the Editor of the StraTechos website.