Sarvam AI Unveils Groundbreaking Large Language Models Surpassing Industry Giants
Bengaluru-based startup Sarvam AI is making waves in the AI industry with the launch of two innovative large language models (LLMs). With a focus on efficiency and scale, Sarvam AI’s latest models are designed to outperform existing giants like DeepSeek’s R1 and Google’s Gemini Flash. This article delves into the details of these models and their potential impact on the AI landscape.
Sarvam AI’s New Large Language Models
Sarvam AI has introduced a 30-billion-parameter model and a 105-billion-parameter model, both built from scratch using a mixture-of-experts (MoE) architecture. This architecture enhances efficiency by activating only a subset of parameters for each input, optimizing performance without the need for the entire model to be engaged.
Pratyush Kumar, Sarvam AI’s Co-founder, emphasized that the 30B model, despite its vast parameter count, activates only 1 billion parameters per token. This approach significantly reduces inference costs and boosts efficiency, particularly for reasoning tasks. The model supports a 32,000-token context window and was trained on an impressive 16 trillion tokens.
The 105B Model: A New Benchmark
The 105-billion-parameter model is a game-changer, activating 9 billion parameters and supporting a 128,000-token context window. Designed for complex reasoning and agentic use cases, this model surpasses DeepSeek’s R1 and Google’s Gemini Flash on key benchmarks. It also excels in Indian language tasks, outperforming Gemini 2.5 Flash.
Kumar highlighted the importance of efficiency in Sarvam’s approach, aiming to deliver AI solutions with a population-scale impact. The company’s focus on efficiency is evident in its strategic use of MoE architecture, which balances scale with cost-effectiveness.
Backing from IndiaAI Mission
Sarvam AI’s advancements come amid the government-backed IndiaAI Mission, which supports sovereign foundational models. With a substantial Rs 10,000-crore fund, the mission has allocated Rs 111 crore for GPU subsidies. Sarvam has already received nearly Rs 99 crore in subsidies and secured 4,096 NVIDIA H100 SXM GPUs through Yotta Data Services, making it the largest beneficiary to date.
The company was the first startup chosen under the mission to build India’s foundational AI model, a testament to its pioneering work in the field.
Expanding Horizons: From AI Models to Smart Eyewear
Beyond developing AI models, Sarvam AI is venturing into hardware with plans to launch ‘Sarvam Kaze’, a smart eyewear product, by May. This device, designed and built in India, will allow developers to build applications on top of it, contributing to the company’s goal of achieving technological sovereignty across various layers, including user-facing devices.
Strategic Partnerships and Collaborations
At the India AI Impact Summit in Delhi, Sarvam AI announced several strategic partnerships. Collaborations with HMD aim to bring AI capabilities to feature phones, while partnerships with Qualcomm and Bosch focus on deploying generative AI solutions across smartphones, PCs, wearables, XR, IoT, automotive, and data centers.
These partnerships highlight Sarvam AI’s commitment to integrating AI into everyday technologies, enhancing functionality and user experience across multiple platforms.
The Future of AI in India
Sarvam AI’s innovations are not just technological feats; they represent a significant step forward in India’s AI journey. By focusing on efficiency, scalability, and strategic partnerships, Sarvam AI is poised to lead the charge in creating impactful AI solutions that cater to diverse needs.
As AI continues to evolve, Sarvam AI’s models set a new standard for performance and cost-effectiveness. Their approach offers a glimpse into the future of AI, where technology is not only advanced but also accessible and practical for a wide range of applications.
Conclusion
Sarvam AI’s new large language models are a testament to the startup’s innovative spirit and commitment to excellence. By surpassing industry leaders on key benchmarks, Sarvam AI is setting a new standard for AI development in India and beyond.
For more information on Sarvam AI and its groundbreaking work, visit Sarvam AI.
This article provides insights into Sarvam AI’s latest achievements, emphasizing the company’s role in shaping the future of AI in India. By focusing on efficiency, strategic partnerships, and technological sovereignty, Sarvam AI is not only advancing AI technology but also contributing to India’s position as a leader in the AI space.







