All you need to know about Sarvam AI, the AI Summit ‘rockstar’

Sarvam AI, a Bengaluru-based artificial intelligence startup founded in 2023 by Dr Vivek Raghavan and Dr Pratyush Kumar, has emerged as the face of India’s push for sovereign AI. With a mission to build a full-stack AI ecosystem tailored to India’s linguistic and cultural diversity, the company is now at the forefront of the country’s technological self-reliance drive.

The startup grabbed national attention at the India AI Impact Summit 2026 in New Delhi, where its pavilion reportedly drew record crowds at Bharat Mandapam. Organisers extended the expo by a day following an overwhelming turnout, with thousands of students, developers, and professionals queuing up to witness live demonstrations of Sarvam’s indigenous AI models.

Why Sarvam AI is in the news

Sarvam is among 12 companies selected under the government’s Rs 10,372 crore IndiaAI Mission to build sovereign language models. It has been entrusted with developing India’s first indigenous large language model (LLM), receiving access to 4,000 high-end GPUs under the programme.

At the summit, the company unveiled two new models — Sarvam 30B and Sarvam 105B — designed for large-scale and advanced reasoning tasks respectively.

The 30B model, built on a mixture-of-experts architecture with one billion activated parameters and pre-trained from scratch on 16 trillion tokens of text, is optimised for efficiency and can power conversational AI across Indian languages, even running on low-end devices such as feature phones.

Also Read: Feb 19 News Live: India built something no other country did: Macron at AI Summit

“Even in the mixture-of-experts world, we have gone towards systems, which can be very, very efficient for inferencing. We think of the 30-billion parameter model as the efficient work horse that will run at scale to power a billion conversations a day and we are looking forward to deploying that across all the things that we do,” said Dr Pratyush Kumar while presenting the model.

The more powerful 105B model, also built on a mixture-of-experts architecture, is aimed at more complex reasoning tasks, with the company claiming benchmark performance competitive with leading global systems.

It uses nine billion activated parameters and is estimated to be roughly twice as expensive as the 30B model.

“It has a larger context window of 128,000 tokens, therefore allows you to do more thinking and do more complex tasks. It is on par with most other open models and with several closed sources frontier models of its class,” Kumar said during the presentation.

Also Read: Bill Gates to skip India AI Impact Summit keynote address

Sarvam shared benchmark comparisons of its reasoning performance against other models, stating that on MMLU-Pro — an advanced AI evaluation benchmark — its model outperformed GPT-120B.

Sarvam’s rise has also been backed by strong investor confidence. In December 2023, it raised USD 41 million in Series A funding from Lightspeed Venture Partners, Peak XV Partners, and Khosla Ventures.

Sarvam-M, Edge, Kaze, Akshar: Indigenous AI products

Sarvam AI’s product stack spans language, speech, vision, and hardware solutions designed for Indian realities.

Its flagship multilingual model, Sarvam-M, is a 24-billion parameter open-weights model fluent in Indian languages. Complementing it are specialised speech models — Bulbul for text-to-speech and Saaras for speech-to-text — supporting over 10 Indian languages.

Sarvam Edge enables on-device AI processing for speech recognition, translation, and synthesis without internet connectivity, addressing privacy and infrastructure gaps.

Also Read: Galgotias’ robodog row: Blaming, philosophy, clarification, but no apology

Sarvam Studio offers end-to-end content transformation, including video dubbing and document translation, while Sarvam Akshar focuses on digitising complex documents and historical manuscripts.

The showstopper at the summit was “Sarvam Kaze”, an indigenous AI-powered smart glasses device capable of real-time visual and auditory processing. The wearable was demonstrated publicly, signalling India’s ambitions not just in AI software but also in hardware innovation.

Building a sovereign AI ecosystem

Sarvam has also signed MoUs with Odisha and Tamil Nadu to establish AI capacity hubs and research parks, reinforcing its state-level partnerships.

By focusing on vernacular AI, code-switching models that understand “Hinglish”, and scalable enterprise tools, Sarvam AI is positioning itself as more than a startup — it is becoming a symbol of India’s ambition to build and control its own AI stack.

Also Read: India AI Summit: 10 things that Google’s Sundar Pichai said

As India accelerates its digital transformation, Sarvam AI’s rise underscores a larger shift: from being a consumer of global AI to becoming a creator of indigenous intelligence.