Sarvam AI redesigns brand amid claims of beating Gemini, ChatGPT on key benchmarks

Sarvam AI has unveiled a redesigned brand identity, introducing what it calls “the all-new Sarvam” and positioning itself as a platform “Designed for all of India.”

In a detailed note accompanying the announcement, the company said the refreshed identity reflects its vision of “AI for all, from India.” Sarvam — derived from the Sanskrit word for “all” — sits at the core of its ambition to build AI that mirrors “how this country thinks, speaks, reasons, and solves problems,” while operating at global standards of technical excellence.

The design centres on the idea of a “gateway,” described as a threshold where “human and machine” and “culture and computation” converge. Inspired by mandala construction, the monogram uses geometric repetition to create a layered, lotus-like form, while a blue-to-orange gradient expresses motion and progression. The overall system, the company said, is intended to feel “grounded, familiar, and modern at the same time.”

The redesign drew strong reactions on X, with users praising the visual identity as “clean” and distinctly Indian. Several commenters said the geometric forms appeared inspired by mandalas and the arched windows of Rajasthan’s forts, calling it “Indic design done right.”

At the same time, many users asked when Sarvam would launch a consumer-facing app. The company's current offerings are positioned as APIs and platforms for developers, businesses and institutions, which may explain the absence of a public-facing app at this stage.

The announcement marks a branding refresh that brings together the company’s recent product launches under a single, sharper narrative focused on India-first artificial intelligence.

The Bengaluru-based startup had already made headlines this week with the launch of two major AI tools — Sarvam Vision and Bulbul V3 — claiming to outperform international AI chatbots such as Google Gemini and OpenAI’s ChatGPT on select benchmarks.

With the new announcement, the company appears to be consolidating those innovations within a broader platform identity.

Sarvam Vision, the company’s document-reading and OCR model, is designed to extract text from images, scanned documents and handwritten notes. Beyond basic text extraction, it can pull data points from graphs, interpret trends from charts and preserve complex table structures, even when tables are nested or visually complicated.

The model supports 22 languages, including major Indian languages such as Hindi, Bengali, Tamil, Telugu and Marathi.

According to co-founder Pratyush Kumar, Sarvam Vision achieved 84.3% accuracy on the olmOCR-Bench, surpassing Gemini 3 Pro at 80.20%, DeepSeek OCR v2 at 78.80% and ChatGPT at 69.80%. It also scored 93.28% on OmniDocBench v1.5, compared to 91.6% for Gemini 3 Pro and 86.56% for ChatGPT 5.2. In word accuracy tests, Sarvam Vision recorded 87.36%, higher than Gemini 3 Pro at 82.51% and ChatGPT 5.2 at 38.60%.

Sarvam AI launched Bulbul V3, a next-generation text-to-speech model built for Indian languages.

The tool offers over 35 high-quality voices across 11 Indian languages and is designed to handle real-world usage patterns such as code-mixing, regional accents, names, abbreviations and emotional tone.

As the company noted in an earlier communication, people in India often switch languages mid-sentence, and accents vary widely by region — requiring voice systems that can adapt without breaking flow.

Bulbul V3 has been tested in independent listening studies and scored strongly on naturalness, robustness and stability.

The company said the model focuses on pacing, emphasis and emotional tone, ensuring that speech feels natural from the first few seconds rather than sounding mechanically generated.

The redesigned Sarvam brand ties these offerings together under a unified message: building AI that is tailored to India’s linguistic diversity and usage patterns.

The evolution also reflects the vision of co-founders Pratyush Kumar and Vivek Raghavan, who have positioned the company around creating AI systems that work reliably across Indian languages at scale.

Also read: From global tech CEOs to business leaders: Big names at India AI Impact Summit 2026