Sarvam AI updates vision model, doubles down on Indic-language push
The update improves document-reading accuracy and focuses on regional scripts used in India. The model outperforms global benchmarks and is tailored for local needs.
Sarvam AI has updated its vision model to enhance document-reading accuracy and further its focus on Indic languages. The company's latest version, Sarvam Vision 2.1, addresses challenges in processing handwritten and scanned documents, particularly those written in the diverse regional scripts used across India. This update is part of Sarvam's broader initiative to develop AI systems that are homegrown and tailored to the Indian market, reducing reliance on foreign technologies.
The company launched the original vision model in February as part of its efforts to build sovereign AI systems for India. The model was designed to read and convert scanned documents and images into usable digital formats, a critical need for businesses dealing with forms, tables, and handwritten records. The updated version, 2.1, incorporates a mix of real-world and artificially generated data, including handwritten forms in multiple Indian languages, to improve performance and usability.
Sarvam Vision 2.1 leads or ranks near the top in several industry benchmarks that test document-reading accuracy, outperforming global competitors such as Gemini 3.6 Flash and Claude Opus 5. The model's performance is particularly notable in handling the complexities of Indic scripts, which are not commonly supported by international AI systems. This achievement positions Sarvam AI as a key player in the regional AI landscape.
In India, where the use of multiple regional scripts is widespread, Sarvam's Indic-language push is expected to have a significant impact. The company has collaborated with IIT Madras's AI4Bharat initiative to enhance its capabilities in processing Indic languages. This focus on local languages is likely to benefit Indian businesses by making AI tools more accessible and effective for their specific needs, potentially reducing costs and improving efficiency in document digitisation.
The updated vision model is expected to influence the broader AI landscape in India, encouraging other companies to invest in region-specific AI solutions. As the model continues to gain traction, it may also prompt discussions around data governance and the need for local AI regulations. With the Indian market showing growing interest in homegrown AI technologies, Sarvam's efforts could set a precedent for future developments in the sector.