Google CEO Sundar Pichai Lauds Sarvam AI's Innovations at India AI Summit
Google CEO Sundar Pichai expressed strong admiration for the work of Sarvam AI during his address at the ongoing India AI Impact Summit 2026. He emphasized the exceptional developer energy in India, stating, "The developer energy I find in India every time I travel, it's bar none, second to none," and praised the country's thriving entrepreneurship ecosystem.
Sarvam AI's Breakthrough in Local AI Models
Pichai specifically highlighted Sarvam AI for its efforts in creating AI models designed for Indian languages and contexts. He remarked, "The work Sarvam has done developing local AI models... I just don't see any impediments to that, and I think it is very, very well positioned." This endorsement comes as Sarvam AI has gained significant attention online, with the startup claiming its AI model outperforms major competitors like Google's Gemini and OpenAI's ChatGPT.
In a notable achievement, Sarvam AI's CEO Pratyush Kumar announced that Sarvam Vision achieved a state-of-the-art accuracy of 84.3% on the olmOCR-Bench English subset, surpassing frontier models such as Gemini 3 Pro and recent OCR models like DeepSeek OCR 2.
What is Sarvam AI?
Founded in August 2023 by Vivek Raghavan and Pratyush Kumar, Sarvam AI focuses on developing AI solutions tailored to Indian needs. The company's model excels in various visual understanding tasks, including:
- Image captioning
- Scene text recognition
- Chart interpretation
- Complex table parsing
A primary goal of Sarvam AI is to unlock India's vast knowledge embedded in physical documents, scanned archives, and historical collections. The company addresses the issue of global models often treating Indian languages as secondary, leading to lower accuracy for regional scripts. Sarvam AI's vision-language model is an inference-efficient 3B state-space model, trained on high-quality datasets covering 22 official Indian languages, including financial documents, literature, newspapers, and historic texts.
Key Technical Features of Sarvam AI
Sarvam AI offers a range of advanced features designed for efficiency and accuracy:
- Speech Recognition: Supports 10 Indian languages within a single 74-million-parameter model (approx. 294MB on device), with automatic language identification, processing speech at 8.5x real-time, and a time-to-first-token under 300 milliseconds on a Qualcomm Snapdragon 8 Gen 3 chipset.
- Speech Synthesis: Has a device footprint of about 60 MB with 24 million parameters, achieving a mean character error rate of 0.0173, indicating high accuracy in synthesized speech across languages. It supports custom voice cloning with about one hour of audio data.
- Translation Model: Features 150 million parameters and an on-device footprint of around 334MB, handling bidirectional translation across 110 language pairs, including 10 Indian languages and English, without intermediate language routing.
How Sarvam AI Differs from Gemini and ChatGPT
Sarvam AI distinguishes itself by prioritizing Indian languages, whereas models like Gemini and ChatGPT often treat them as secondary. Trained in 22 Indian languages, Sarvam AI provides higher accuracy for regional scripts. Additionally, it goes beyond text extraction to interpret visual elements, enhancing understanding of complex documents through its large-scale Indic OCR benchmark for Indian languages.
Availability and Key Features
The Document Intelligence API is free for February 2026, allowing users to explore and build with Sarvam Vision at scale. Key features include:
- Multimodal Vision-Language: Understands images and texts together for tasks like image captioning and chart interpretation.
- Document Understanding (Indian Languages Focused): High-accuracy OCR and knowledge extraction for 22 Indian languages, including historic texts.
- Charts and Data Interpretation: Capable of understanding charts, data, illustrations, and visual analysis.
- Multilingual Visual: Interprets visual elements across multiple languages in the same document.
- Leading Performance: Excels in global English benchmarks and introduces the Sarvam Indic OCR Bench for Indian languages.
- Accessible API: Production-ready document intelligence APIs free for experimentation in February 2026.
Sarvam AI's innovations represent a significant step forward in making AI more accessible and effective for Indian users, as highlighted by Sundar Pichai's praise at the summit.
