Sarvam AI just unveiled two indigenous LLMs at India AI Impact Summit 2026—a 30B parameter model for...
The AMW Read
The launch of high-parameter indigenous models by Sarvam AI updates the player map for frontier labs in emerging markets and signals a structural shift toward sovereign AI in India.
Sarvam AI just unveiled two indigenous LLMs at India AI Impact Summit 2026—a 30B parameter model for real-time conversations and a 105B model for complex reasoning, both built from scratch for India's multilingual ecosystem. The 105B model uses Mixture-of-Experts architecture, activating only 9B parameters per token while supporting all 22 official Indian languages with a 128K context window, and reportedly outperforms DeepSeek R1 (600B parameters) on key benchmarks despite being one-sixth the size. This marks a decisive shift toward sovereign AI, reducing India's dependence on foreign models while enabling stronger data governance for domestic enterprise and government applications. With Google announcing subsea fiber expansion and OpenAI deepening academic partnerships, India is rapidly positioning itself as an AI co-creator rather than just a consumer market.
