Bengaluru-based startup Sarvam AI is drawing global attention for its artificial intelligence systems Sarvam Vision and Bulbul V3, after its co-founder and CEO Pratyush Kumar reported strong benchmark performances on optical character recognition (OCR) tasks and outlined the company’s “sovereign AI” ambitions. In posts on X (formerly Twitter) and statements on the firm’s website, Sarvam AI said its models surpassed several established systems on key benchmarks, while expanding speech technology across Indian languages. Tech commentator Deedy Das and KissanAI founder Pratik Desai have also publicly commented on the startup’s progress.

Key Takeaways on Sarvam AI

• Sarvam AI is a Bengaluru-based startup founded in 2023.
• CEO Pratyush Kumar said Sarvam Vision achieved 84.3% accuracy on olmOCR-Bench and 93.28% on OmniDocBench v1.5.
• Bulbul V3 supports 35 voices across Indian languages, with expansion planned to cover all 22 scheduled languages.
• The company described its goal as building “sovereign AI” systems tailored to India’s needs.
• Tech commentator Deedy Das and KissanAI founder Pratik Desai posted assessments on X regarding Sarvam’s products and pricing.

Sarvam AI and Its ‘Sovereign’ Vision

Sarvam AI is a Bengaluru startup founded in 2023 that focuses on building foundational artificial intelligence models developed entirely in India. According to the company’s website, it aims to make AI widely accessible across the country.

“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 wrote.

The startup has positioned its work around what it calls “sovereign AI,” describing this approach as building self-dependent, India-focused systems that aim to match world-class performance while addressing local priorities.

Also Read:  Top Educational Magazines, Software & Homeschool Tools for Kids in 2026

Sarvam AI’s progress has been framed as part of India’s broader advances in artificial intelligence, with the company highlighting potential adoption of its tools across sectors such as banking, education, and government services.

Benchmark Results for Sarvam Vision

Sarvam Vision, the company’s OCR-focused model, has been at the center of recent attention.

In a post on X, Pratyush Kumar said the system achieved:

• 84.3% accuracy on olmOCR-Bench, surpassing Gemini 3 Pro and DeepSeek OCR v2.
• 93.28% overall on OmniDocBench v1.5.

He added, “On Indian languages, Sarvam Vision is the best model by far, while supporting all 22 scheduled Indian languages.”

The startup also said its model performed well on difficult document-processing tasks, including:

• Image captioning
• Scene text recognition
• Chart interpretation
• Complex table parsing

Sarvam AI described Sarvam Vision as part of a series that includes a 3B-parameter state-space vision-language model, capable of visual understanding tasks across a range of document types.

Users were cited as valuing its performance on everyday, real-world documents, particularly in areas where traditional OCR systems often struggle, such as complex layouts, technical tables, and mathematical formulas.

Bulbul V3 and Speech Technology for Indian Languages

Alongside its OCR systems, Sarvam AI has developed Bulbul V3, a text-to-speech model aimed at Indian language use cases.

According to the company’s statements, Bulbul V3:

• Provides 35 distinct voices.
• Supports 11 Indian languages at present.
• Is being expanded to cover all 22 scheduled Indian languages.

In earlier descriptions, the company said its sample sets span historical and modern content, ranging from the year 1800 to the present, and include documents with different scan qualities and formats.

Bulbul V3 has also been compared with global speech-generation systems. The startup said the model competes on quality with ElevenLabs while being priced more affordably for Indic-language applications.

Also Read:  International Day of Education: Know About the Real Aim of Education

KissanAI founder Pratik Desai wrote on X:

“We use Bulbul as our go-to TTS model for our Indic use cases, and they have just gotten better with each release. Meanwhile, ElevenLabs’ cost never made sense for Indic or any other languages.”

Leadership and Background of CEO Pratyush Kumar

Sarvam AI was co-founded by Pratyush Kumar and Vivek Raghavan, with Kumar currently serving as CEO.

Before starting Sarvam AI, Kumar launched:

• AI4 Bharat, focused on Indian-language AI applications.
• PadhAI, an initiative for affordable online learning.

His academic background includes a Ph.D. from ETH Zurich and a B.Tech from IIT Bombay. He has worked at Microsoft Research and IBM Research and serves as adjunct faculty at IIT Madras. Kumar also shares updates about Sarvam AI’s progress on X.

Recognition From Industry Voices

Sarvam AI’s work has drawn responses from technology commentators and startup founders.

Tech commentator Deedy Das, who previously questioned the focus on training smaller Indic-language models, posted a reassessment on X.

“I was wrong about Sarvam,” Das wrote.

He added:

“When I wrote about them a year ago, I felt like the direction to train small ‘indic’ language models was wrong. But boy, have they turned it around. They have the best text-to-speech, speech-to text, and OCR models for Indic languages, and that’s actually really valuable. The pricing is very reasonable. And the website is not only beautifully designed but dirt easy to use.”

Das further wrote:

“They’re filling a well needed gap in the ecosystem and doing things big labs will probably never focus on to the fullest extent (at least in the short term). I don’t know anything about the business, but there’s a lot to appreciate about what they’ve built technologically and I can’t remember the last time I felt that way about software products coming out of India. Well done.”

What Sarvam AI’s Progress Signals for India

Sarvam AI’s reported benchmark scores and language-focused tools have been described as a notable development in India’s artificial intelligence landscape.

The startup said its success highlights the country’s potential in core AI innovation and reflects growing efforts to address India-specific challenges through locally built systems.

By focusing on OCR, speech generation, and multilingual support, the company has positioned its products for possible deployment across public and private sector services, while continuing to frame its mission around domestic technological capability.

FAQs on Sarvam AI