Sarvam AI: Bengaluru’s bid for sovereign AI

Sarvam AI, a Bengaluru startup, is redefining India’s AI with sovereign models. Sarvam Vision beats Gemini, ChatGPT, and Anthropic in OCR benchmarks, while Bulbul V3 delivers natural text-to-speech in 11 Indian languages, expanding soon to 22.

author-image
Archana Reddy
Sarvam AI
Advertisment
  • Bengaluru startup’s OCR tool outperformed Gemini, ChatGPT, and Anthropic
  • New text-to-speech model supports 35 voices across 11 Indian languages
  • Focus on Indic languages fills a gap global AI labs often overlook

Sarvam AI in Bengaluru shines with Sarvam Vision beating global OCR rivals and Bulbul V3 offering natural Indic text-to-speech

India’s AI landscape has long been overshadowed by developments in the US and China. Yet Bengaluru-based startup Sarvam AI is positioning itself as a serious contender by building foundational AI models entirely in India. Branding itself as “sovereign AI,” the company is gaining recognition with two recent tools — Sarvam Vision and Bulbul V3.

Sarvam Vision’s Breakthrough

Sarvam Vision, an optical character recognition (OCR) model, has achieved benchmark scores surpassing global leaders like Google Gemini, Anthropic Claude, and even ChatGPT in its domain. On the olmOCR-Bench, it recorded an accuracy of 84.3 percent, outperforming Gemini 3 Pro and DeepSeek OCR v2. On OmniDoc Bench v1.5, it scored 93.28 percent, excelling in complex layouts, technical tables, and mathematical formulas — areas where traditional OCR systems often falter. This performance has drawn global attention, shifting skepticism about Sarvam’s India-focused approach into approval.

Bulbul V3 Expands Voice AI

Alongside OCR, Sarvam has launched Bulbul V3, a text-to-speech model designed for Indian languages. Supporting over 35 voices across 11 languages, with plans to expand to 22, Bulbul V3 aims to deliver natural, expressive, and production-ready audio. It reduces common errors and ensures accurate, consistent speech, making it particularly valuable for India-specific use cases.

Also Read: Karnataka’s auto land conversion reform: What it means for Bengaluru’s real estate?

Filling a Critical Gap

Sarvam’s focus on Indian languages addresses a gap often overlooked by global AI labs. Its OCR and speech models are tailored to Indic contexts, offering affordability and accessibility compared to international competitors. This approach has earned praise from both users and industry observers, who see Sarvam’s work as a step toward democratizing AI for India’s diverse linguistic landscape.

Outlook

By combining technical excellence with local relevance, Sarvam AI is carving out a niche in the global AI ecosystem. Its success with Sarvam Vision and Bulbul V3 demonstrates that India can produce competitive, world-class AI models while addressing domestic needs. As the company scales, it may redefine perceptions of India’s role in advanced AI development.

Also Read: Devanahalli real estate: Bengaluru’s emerging Gurugram?

AI startup startup capital Bengaluru startups Bengaluru
Advertisment