Bengaluru-based Sarvam AI has launched two foundational models, Sarvam-30B and Sarvam-105B, at the I...
The AMW Read
Updates the player map with a significant non-Western frontier lab and signals the execution of India's sovereign AI strategy via indigenous model development.
NoveltySignificance
Foundation Models · Player MapGeopolitics
Bengaluru-based Sarvam AI has launched two foundational models, Sarvam-30B and Sarvam-105B, at the India AI Impact Summit 2026, marking the first time an Indian startup has created large-scale AI models competing with global giants. The 105B model with 128,000-token context window claims to outperform DeepSeek R1 and Gemini Flash on key benchmarks, while the 30B model supports 22 Indian languages and even works on feature phones. Both models use mixture-of-experts architecture and will be open-sourced, reflecting India's sovereign AI strategy to build indigenous infrastructure. With over $50 million in funding and government GPU support, this signals India's emergence as a serious player in the global AI race.
