India’s AI ecosystem has received a boost with the introduction of two cutting-edge models from Bengaluru-based startup Sarvam: Bulbul V3 and Sarvam Vision. These models have reportedly achieved benchmark-defying results, outperforming established AI models from global leaders such as OpenAI and Google.
Sarvam, co-founded by Pratyush Kumar, has positioned itself as a challenger in the artificial intelligence landscape by focusing on optical character recognition (OCR) tasks. OCR, which enables machines to recognize and process text from images and documents, is a critical technology used in document digitization, data entry automation, and enterprise applications.
According to recent reports shared on social media, Sarvam Vision achieved 84.3 percent accuracy on the olmOCR-Bench, outperforming models like DeelSeek OCR v2 and Google Gemini 3 Pro. The model also scored an impressive 93.82 percent on OmniDocBench v1.5, demonstrating high efficiency in processing diverse document formats. These results suggest that Sarvam Vision not only competes with international AI models but may lead in specific OCR applications.
Bulbul V3, the startup’s other AI offering, complements Sarvam Vision by focusing on natural language processing, image understanding, and cross-modal AI tasks. While Sarvam Vision excels in text recognition, Bulbul V3’s strength lies in understanding and interpreting complex data, which makes the two models highly versatile for enterprise and industrial use cases.
Pratyush Kumar highlighted that Sarvam’s success is a reflection of India’s growing capability in AI research and development. By leveraging local talent and computing resources, Sarvam has been able to create AI models that perform at a global standard. The startup’s approach emphasizes fine-tuning models for specific tasks rather than relying solely on generic large language model architectures.
Experts believe that these developments signal a turning point for India’s AI sector. While international AI giants dominate global benchmarks, Sarvam’s achievements show that homegrown solutions can compete effectively, particularly in specialised tasks like OCR and document analysis. The results also highlight the potential for Indian startups to provide enterprise-grade AI solutions at competitive costs, which may accelerate AI adoption across industries in India.
Industry analysts note that Sarvam’s models could have far-reaching applications beyond OCR. From automating government document processing to enhancing business intelligence, the models are likely to benefit sectors such as banking, legal, healthcare, and education. This success also positions India as a growing hub for AI innovation capable of producing models that rival those of global technology leaders.
As AI adoption increases across sectors in India, startups like Sarvam demonstrate that local innovation can not only match but occasionally surpass international benchmarks. With Bulbul V3 and Sarvam Vision leading the charge, India is poised to make significant contributions to the global AI landscape in the coming years.
