Sarvam AI Outperforms Global AI in India

Bengaluru-based Sarvam AI has achieved significant breakthroughs with its Sarvam Vision OCR tool and Bulbul V3 voice model, outperforming leading global AI systems like Google===Google Gemini and OpenAI===ChatGPT in Indian-language tasks. This marks a major milestone for India's emerging sovereign AI ecosystem, demonstrating that homegrown firms can develop world-class foundational models. Sarvam Vision achieved 84.3% accuracy on olmOCR-Bench and 93.28% on OmniDocBench v1.5, excelling in complex Indian document layouts. Bulbul V3 supports over 35 voices across 11 Indian languages, offering natural and expressive speech. These achievements highlight Sarvam AI's focus on India-specific challenges, such as multilingual documentation and regional language use, which global platforms often overlook. The company's success is seen as a boost for India's AI ambitions, reducing reliance on foreign AI platforms and promising improved digital access for governance, banking, and education within the country. Co-founded by Vivek Raghavan and Pratyush Kumar, Sarvam AI's approach emphasizes efficiency and local data curation, proving that smaller, optimized models can deliver competitive performance.