by Sarvam AI, Bangalore
Indian AI startup building foundation models optimised for Indian languages and contexts. Launched Vikram and 105B-parameter models at India AI Impact Summit 2026. Focused on enabling Indian businesses to build AI applications with models that understand Indian languages, culture, and business contexts natively.
Foundation models (Vikram series) trained on Indian language data. 105B-parameter model for enterprise-grade Indian language understanding. APIs for text generation, translation, and summarisation in Indian languages. Models optimised for Indian business contexts (legal, finance, government). On-premise deployment options for data sovereignty. Building blocks for enterprises to create custom AI agents.
Indian enterprises wanting AI models that natively understand Hindi, Tamil, Telugu, and other Indian languages. Government and public sector organisations needing sovereign AI solutions. Developers building India-specific AI applications.
Newer company — enterprise track record still developing. Model ecosystem smaller than OpenAI/Anthropic/Google. Tooling and developer experience still maturing. Limited third-party benchmarks available.
Compliance status based on publicly available information. AgentVault does not independently certify compliance. Verify directly with the provider. Read our scoring methodology.