MUMBAI, Maharashtra — Bengaluru-based startup Sarvam AI said its latest vision and speech models have outperformed larger global competitors on key optical character recognition and text-to-speech benchmarks, particularly for Indian languages.
In a post on social media, Sarvam AI co-founder Pratyush Kumar said the company’s newest OCR model delivered state-of-the-art performance on widely used benchmarks. “Sarvam Vision achieves state-of-the-art accuracy of 84.3 per cent on the olmOCR-Bench (English only subset) outperforming frontier models like Gemini 3 Pro and recent OCR models like DeepSeek OCR 2,” Kumar said.
According to Kumar, Sarvam Vision also posted a 93.28 per cent overall score on OmniDocBench v1.5 for the English-only subset, performing strongly on complex formulas and layout parsing and coming close to the current best-in-class results.
He added that the company’s Bulbul V3 text-to-speech model supports 35 voices across all 22 scheduled Indian languages and is capable of handling documents with varying scan quality and content types. “On Indian languages, Sarvam Vision is the best model by far, while supporting all 22 scheduled Indian languages,” Kumar said.
The Vision series includes a three-billion-parameter state-space model designed to handle image captioning, scene text recognition, chart interpretation and complex table parsing. In examples shared online, Sarvam Vision was shown accurately extracting technical terminology from dense tables with merged rows and columns, as well as pulling structured data from charts published in the latest Economic Survey.
The model has also demonstrated broader scene understanding, with examples showing it correctly describing natural landscape photographs.
Sarvam AI said its broader goal is to make artificial intelligence widely accessible across India. “We want India to embrace the most important technological shift of our time with confidence and control. Our ambition is to build foundational components and apply them to the country’s unique needs,” the company said.
Union IT Minister Ashwini Vaishnaw said in a recent post that the startup’s progress reflects the growing impact of India’s national push to build domestic artificial intelligence capabilities. (Source: IANS)
