Sarvam AI slashes Vision API pricing by two-thirds after processing 35 million pages

As businesses race to digitise everything from government records and legal archives to invoices and handwritten forms, one Indian AI startup is betting that lower prices could accelerate adoption even further.

Sarvam AI has announced a steep reduction in pricing for Sarvam Vision, its document intelligence platform, after reporting strong uptake among developers and enterprise customers. The move comes just months after the product's launch and signals the company's growing confidence in both the technology and the market demand for AI-powered document processing.

The Bengaluru-based startup says more than 35 million pages have already been digitised through the platform, a milestone that has helped it improve efficiency and significantly reduce operating costs.

Adoption surge drives major price cut

Sarvam Vision was introduced earlier this year as a vision-language model designed to help organisations extract, understand and digitise information from documents. The platform combines optical character recognition (OCR) capabilities with AI-powered document understanding, enabling users to process everything from scanned records and forms to complex multi-page files.

More from Tech

In a series of posts on X, the company revealed that it is reducing the price of the Sarvam Vision API from Rs 1.5 per page to Rs 0.5 per page, representing a reduction of roughly 67 per cent.

The decision follows a sharp increase in usage. According to Sarvam, developers and partners have collectively processed more than 35 million pages through the platform since its debut.

Rather than simply benefiting from higher volumes, the company says it used the growth in demand as an opportunity to overhaul key parts of its infrastructure. As workloads increased, Sarvam re-engineered elements of its serving architecture to improve efficiency at scale, ultimately lowering the cost of delivering the service.

The company says these gains stem from improvements in inference performance, smarter batching techniques and better utilisation of computing resources across its sovereign cloud infrastructure.

For organisations handling large document collections, the lower pricing could substantially reduce the cost of digitisation projects, making AI-powered document processing more accessible to businesses of varying sizes.

Building AI for India's document-heavy economy

Sarvam Vision is designed with India's multilingual environment in mind. The platform supports all 22 officially recognised Indian languages, allowing organisations to process documents across a wide range of scripts and formats.

The model is capable of handling scanned paperwork, handwritten records, forms, financial documents, archives and other content that often proves difficult for conventional OCR systems. By combining visual understanding with language capabilities, the system aims to extract information more accurately from complex layouts and diverse document types.

The company has previously highlighted strong benchmark performance for the model, reporting competitive results on document understanding evaluations such as olmOCR-Bench and OmniDocBench.

The latest pricing move also reflects a broader strategy. Sarvam has positioned itself as a sovereign AI company focused on building foundational AI technologies tailored to India's needs while running on domestic infrastructure.

Demand for document intelligence tools is rising across sectors including banking, healthcare, education and government services, where large volumes of paper-based information still need to be digitised and analysed.

By lowering costs while expanding capacity, Sarvam appears to be pursuing a familiar technology playbook: use scale-driven efficiencies to attract more users, drive broader adoption and strengthen its position in India's increasingly competitive AI ecosystem.