Sarvam AI is an Indian artificial intelligence company headquartered in Bengaluru, Karnataka. Founded in August 2023, it develops a full-stack sovereign AI platform focused on India’s linguistic diversity, including large language models (LLMs), speech systems, vision-language models, and enterprise AI solutions supporting 22 officially recognized Indian languages and code-mixed text.
The company has emerged as a key player in India’s sovereign AI initiatives under the IndiaAI Mission, emphasizing data sovereignty, on-premises deployment, and multilingual accessibility. Its flagship consumer product is the Indus AI assistant. In June 2026, Sarvam AI achieved unicorn status after raising $234 million in its Series B funding round at a $1.5 billion valuation.
History
Sarvam AI was founded in August 2023 by Dr. Vivek Raghavan and Dr. Pratyush Kumar, both formerly associated with AI4Bharat at IIT Madras. Vivek Raghavan previously contributed to India’s digital public infrastructure, including Aadhaar and UPI, while Pratyush Kumar led open-source AI efforts for Indian languages.
In December 2023, the company raised $41 million in a combined seed and Series A round, led by Lightspeed Venture Partners with participation from Peak XV Partners and Khosla Ventures. This was one of the largest early-stage AI funding rounds in India at the time.
In April 2025, the Ministry of Electronics and Information Technology (MeitY) selected Sarvam AI under the IndiaAI Mission to develop indigenous foundational models, providing access to government-supported computing resources.
In February 2026, Sarvam unveiled its flagship models at the India AI Impact Summit in New Delhi. It open-sourced key models in March 2026.
On 15 June 2026, Sarvam announced the first close of its Series B round, raising $234 million at a $1.5 billion post-money valuation (part of a targeted $300 million round), led by HCLTech with participation from Bessemer Venture Partners and existing investors. This made Sarvam India’s newest AI unicorn.
Corporate structure and finances
Sarvam AI operates as a privately held company building a complete sovereign AI stack, including models, inference infrastructure, and deployment options (cloud, private cloud, on-premises, and air-gapped environments). It prioritizes enterprise-grade security, data residency, and compliance with Indian regulations such as DPDP.
As of June 2026, the company has raised over $275 million cumulatively and maintains a strong focus on reducing dependence on foreign AI infrastructure.
Products and technology
Sarvam develops models trained from scratch on Indian-language datasets, using architectures such as Mixture-of-Experts (MoE) for efficiency. Its systems emphasize voice-first interactions, cultural relevance, and real-world Indian use cases (e.g., government forms, agriculture, and BFSI).
Foundational language models
- Sarvam-30B (February 2026): A 30-billion parameter MoE model (activating ~2.4 billion parameters per token) optimized for efficiency, reasoning, coding, and multilingual tasks with a 32K context window. Suitable for on-device and resource-constrained environments.
- Sarvam-105B (February 2026): The flagship 105-billion parameter MoE model (activating ~10.3 billion parameters per token) with a 128K context window. It excels in complex reasoning, agentic workflows, mathematics, and enterprise applications.
Indus (launched 20 February 2026): A consumer-facing AI chatbot/assistant powered by the Sarvam-105B model. Available on web, iOS, and Android, it supports voice and text interactions across Indian languages and serves as a locally relevant alternative to global chatbots.
Earlier models include Sarvam-1 (2024) and Sarvam-M (2025).
Speech and multimodal systems
- Saaras: Advanced speech-to-text (ASR) system supporting up to 23 languages.
- Bulbul: Natural text-to-speech synthesis with support for multiple Indian languages and accents.
- Sarvam Vision / Mayura: Vision-language models for document understanding, OCR in Indian scripts, and structured content processing.
- Sarvam Translate: High-quality multilingual translation supporting formal and colloquial usage.
Devices and enterprise tools
- Sarvam Kaze: An upcoming AI-powered wearable glasses device for real-time voice interaction, vision capture, and translation (planned for 2026).
- Enterprise platform including voice agents, workflows, and “Arya” for building production-grade AI applications with forward-deployed support.
Open-source contributions
In March 2026, Sarvam open-sourced the Sarvam-30B and Sarvam-105B models under the Apache 2.0 license. Model weights are available on Hugging Face (under the sarvamai organization) and AIKosh, promoting accessibility for developers and enterprises.
Government and institutional collaborations
Sarvam AI participates in the IndiaAI Mission and has partnered with the Unique Identification Authority of India (UIDAI) to integrate voice-based and multilingual AI capabilities into Aadhaar services, deployed on-premises for data security.
It collaborates with academic institutions via AI4Bharat and has enterprise/government deployments in sectors including agriculture, insurance, BFSI, and public services. Additional partnerships include Pixxel (orbital data center), HCLTech, and others for compute and deployment.
In March 2026, Sarvam launched a Startup Program offering API credits, priority engineering support, and infrastructure to early-stage companies building multilingual AI applications.
Reception
Sarvam’s models are recognized for strong performance on Indian-language benchmarks, document understanding, and cultural relevance, often competing with or outperforming global models in region-specific tasks. The company is viewed as a cornerstone of India’s push for digital and AI sovereignty.
See also
- Artificial intelligence in India
- IndiaAI Mission
- List of large language models
- Open-source artificial intelligence