Sarvam AI

Sarvam is India's full-stack sovereign AI platform, offering speech-to-text, text-to-speech, translation, and conversational agents across 22 Indian languages. In February 2026, the company unveiled two foundational large language models at the AI Impact Summit in New Delhi: Sarvam-30B, a mixture-of-experts model, and Sarvam-105B, a 105-billion parameter model with a 128,000-token context window positioned for complex reasoning and enterprise applications. Both models were subsequently open-sourced under the Apache License. Sarvam's models have outperformed GPT-4o, Gemini, and Llama 70B on Indian language benchmarks covering OCR, document layout understanding, and Indic language processing. The platform also includes Sarvam Vision for document understanding and a voice AI stack that powers real-time conversational agents in Indian languages. Founded in 2023 and headquartered in Bengaluru, Sarvam AI was among 12 organisations selected by the Indian government to develop AI models built on Indian datasets. The company is led by CEO Pratyush Kumar, an IIT Bombay alumnus with a PhD in computer engineering and prior research roles at IBM and Microsoft, and co-founder Vivek Raghavan, an IIT Delhi and Carnegie Mellon alumnus who spent nearly 12 years working on Aadhaar's biometric systems at UIDAI. Backed by Lightspeed Venture Partners, Peak XV Partners, and Khosla Ventures with $41 million raised, Sarvam is building the AI stack India can call its own, trained on Indian data, hosted on Indian compute, and designed for India's 1.4 billion people.