India’s Sarvam AI Outshines Google Gemini And ChatGPT In Indic Language OCR Breakthrough
Bengaluru : Sarvam AI, a Bengaluru-based startup, is emerging as a key player in India’s push for technological self-reliance with its latest AI advancements. The company’s models, particularly Sarvam Vision and Bulbul V3, have demonstrated superior performance in specialized tasks, drawing attention both domestically and internationally.
Sarvam Vision, a 3-billion-parameter vision-language model built on state-space architecture, has excelled in optical character recognition (OCR), especially for Indian languages.
On OmniDocBench v1.5 (English only subset), Sarvam Vision achieves 93.28% overall score, excelling in complex formulas and layout parsing and being within touching distance of the current state of the art. pic.twitter.com/7YDfbX1pCz
— Pratyush Kumar (@pratykumar) February 5, 2026
According to claims shared by co-founder Pratyush Kumar on X (formerly Twitter), the model recorded 84.3% accuracy on the olmOCR-Bench benchmark, outperforming Google’s Gemini 3 Pro and DeepSeek OCR v2. It also achieved 93.28% on OmniDocBench v1.5, which evaluates performance on real-world documents including complex layouts, technical tables, and mathematical content. Kumar highlighted that Sarvam Vision stands out as the leading model for Indian languages while supporting all 22 scheduled languages.
In parallel, Bulbul V3, the company’s text-to-speech model, offers 35 distinct voices distributed across those same 22 official Indian languages, enabling high-quality speech generation tailored to regional needs.
Sarvam AI positions itself as a “sovereign” AI initiative, focused on creating foundational AI components that address India’s specific requirements and promote widespread accessibility. As stated on its official website, the company seeks to ensure that India adopts this transformative technology with greater confidence and control, adapting it to local contexts.
The developments have garnered praise from global observers, including tech commentator Deedy Das, who previously expressed skepticism about small Indic language models but later commended Sarvam AI. Das noted on X that the startup now delivers the best text-to-speech, speech-to-text, and OCR capabilities for Indic languages, with affordable pricing and an intuitive, well-designed platform. He emphasized the importance of filling ecosystem gaps that larger global labs may overlook in the near term.
These achievements from Sarvam AI represent a notable step forward in India’s AI landscape, potentially accelerating the integration of tailored AI tools in sectors such as banking, education, and government services.
Also read : UK Prime Minister’s Chief Of Staff Morgan McSweeney Steps Down Amid Epstein-Mandelson Controversy