Sarvam AI, a Bengaluru-based startup, has unveiled plans for a trillion-plus parameter model, launched an India-hosted inference platform, opened a San Francisco research office, and reportedly raised another $75 million in a funding round led by Nvidia Corp., just months after securing $234 million. The company plans to use the funds for next-generation AI models for applications such as AI agents, software development, and cybersecurity, while expanding computing capacity. Sarvam’s India-hosted inference platform serves its own models and others like GLM 5.2 and Gemma 4, signaling a strategy to become an AI platform. Sarvam says its models are built from scratch rather than distilled from existing open AI models. Sarvam 105B performs well on reasoning, programming, and agentic tasks, while Sarvam 30B is optimized for real-time deployment. Both models score high on Indian language benchmarks, outperforming models from companies including OpenAI, Anthropic, Google, Alibaba (Qwen), High-Flyer (Deepseek), Z.ai (GLM), and Moonshot AI (Kimi). The models are built to balance capability with efficiency, keeping computing and inference costs under control.