Top News

Sarvam Vision OCR: Till now whenever Artificial Intelligence i.e. AI models Whenever there was talk of India, the names of America and China used to come up first. But now the picture seems to be changing. Bengaluru’s homegrown startups Universal AI has achieved such a feat that it has surprised even the world’s biggest AI models. The special thing is that this company has been preparing foundational AI models in India since the beginning. Recently, two tools of Sarvam AI, Sarvam Vision and Bulbul, have created a stir in the tech world.

What is Sarvam Vision AI?

Sarvam Vision is an advanced AI system that has demonstrated excellent performance in the field of Optical Character Recognition (OCR). This homegrown AI has given better results than popular models like ChatGPT, Google Gemini and Anthropic Claude on some international benchmarks. This is the reason why now not only users but also tech experts are praising it. Sarvam AI co-founder Pratyush Kumar has shared the achievements of in-house AI models through several posts on the X platform.

Victory of indigenous AI on global benchmark

According to the founder’s post, Sarvam Vision achieved an accuracy of 93.28 percent on OmniDocBench v1.5 (English only subset), where models like Gemini and ChatGPT lagged behind.

This benchmark tests how well an AI system can read and understand real-world documents. Sarvam Vision also scored well in sections like complex design, technical tables and math formulas where common OCR systems often fail.

Old questions turned into compliments

There was a time when Sarvam AI had to face questions about focusing only on Indian languages, but now the same questions have turned into praise. Tech commentator DD Das on

Users are also appreciating

Users are also sharing their experiences regarding Sarvam AI. One user wrote that he used this tool a few days ago and the experience was “amazing” Remained.

Bulbul V3: Voice of AI in Indian languages

Along with OCR, Sarvam AI has also launched its new text-to-speech model Bulbul V3. The company said in a blog post, ‘We are launching Bulbul V3. This is the company’s most capable text to speech model” This model is designed to generate natural, expressive and production-ready voices for Indian languages. Bulbul V3 supports 11 Indian languages ​​and more than 35 voice options. The company says that more languages ​​will be added to it in the coming time.