In the fast-evolving world of artificial intelligence, most conversations are dominated by American and Chinese giants. But in India, a new player is positioning itself as the architect of “Sovereign AI”—AI built from the ground up for the country’s unique needs. That player is Sarvam AI.
This isn’t just another startup cloning ChatGPT. Sarvam AI is attempting to solve a fundamentally different problem: how to make AI work for a population that speaks 22 languages, uses voice more than text, and requires infrastructure that is both affordable and locally controlled.
In this guide, we will cut through the hype to explain what Sarvam AI actually is, who is behind it, how to access its tools, and whether it’s a viable option for developers, job seekers, and businesses. We’ll also address the controversies and questions you’re searching for, from sarvam ai funding to is sarvam ai any good.
What is Sarvam AI?
Founded in August 2023 and headquartered in Bengaluru, Sarvam AI is a research and product startup building a full-stack AI ecosystem for India. Its mission is to create foundational AI models that reflect India’s cultural and linguistic context.
Unlike global models trained primarily on English internet data, Sarvam trains its models on “trillions of tokens” of Indian-language data, including Hinglish and other mixed dialects . Their core differentiator is a “voice-first” approach. The company believes that in a country with varying literacy levels, the primary interface for AI will be speech, not text.
Their offerings fall into three buckets:
- Foundation Models: Large Language Models (LLMs) like the recently open-sourced Sarvam 30B and 105B models .
- Enterprise APIs: Tools for businesses to integrate Indic-language AI, such as the Sarvam Vision OCR and Bulbul V3 text-to-speech .
- Consumer Apps: The Indus app, a chatbot available to the public .
Founders and Background
To understand sarvam ai founder dynamics, you need to look at the deep public-sector tech experience at the helm. The company was started by two veterans who previously collaborated at AI4Bharat, a research centre at IIT Madras .
- Vivek Raghavan: He is the seasoned infrastructure builder. An IIT Delhi and Carnegie Mellon alumnus, he spent over a decade as a volunteer technology advisor for the Aadhaar project—India’s unique identity system that now covers over a billion people. He understands how to build and deploy technology at a population scale. His focus is on using AI for social impact, having also worked with the Supreme Court of India on translation software (SUVAS) .
- Pratyush Kumar: He is the AI research expert. With a PhD from ETH Zurich and stints at IBM Research and Microsoft Research, he brings academic rigour. As an adjunct faculty at IIT Madras and co-founder of AI4Bharat, he has been at the forefront of creating open-source language datasets and models for India .
Together, they combine deep tech stack knowledge with real-world deployment experience, which is a rare combination in the Indian startup ecosystem.
Funding and Backers
Who funds Sarvam AI? The list is a who’s who of Silicon Valley and Indian venture capital.
In December 2023, Sarvam AI announced a massive $41 million funding round, combining Seed and Series A. This was one of the largest early-stage cheques for an Indian AI startup at the time .
- Lead Investors: Lightspeed, Peak XV Partners (formerly Sequoia India), and Khosla Ventures.
- Strategic Backing: Notably, Vinod Khosla (co-founder of Sun Microsystems) personally backed the company, signalling confidence in the team’s vision .
sarvam ai funding also includes a unique government component. In 2025, it was selected under the IndiaAI Mission to build a sovereign large language model. In exchange for equity, the government provided subsidised access to 4,096 NVIDIA H100 GPUs at the Yotta data centre, valued at nearly ₹25 crore . This public-private partnership is central to their ability to train large-scale models without prohibitive capital costs.
Products, API, and Pricing
Sarvam AI API and Capabilities
For developers and enterprises, the sarvam ai api is the primary gateway. The platform offers several specialized models via API:
- Sarvam Vision: A multimodal model excelling at Optical Character Recognition (OCR) and document understanding. It reportedly scored 84.3% on the olmOCR-Bench, outperforming Gemini 3 Pro and DeepSeek OCR v2 on Indian language scripts .
- Bulbul V3: A text-to-speech model designed for multiple Indian languages and accents .
- Sarvam Arya: A language model designed for conversational AI.
Is Sarvam AI API free?
As of March 2026, there is a free tier for experimentation. Sarvam AI announced that its Document Intelligence API (Sarvam Vision) was free for the month of February 2026 to allow users to build and test at scale . While that specific promotion has ended, startups typically maintain a limited free tier for developers. You would need to check their official website for the current “free trial” quotas.
sarvam ai pricing
Is Sarvam free to use? For basic consumer usage, yes (via the Indus app). For the API, pricing is not static; it is typical for B2B AI APIs to charge based on usage—per token (for language models) or per page/document (for Vision OCR). For enterprise deals involving fine-tuning or dedicated instances, sarvam ai pricing is usually custom and requires contacting their sales team. Related: How to Choose an AI API for Indic Languages
Example API Request (Pseudo-code)
To use Sarvam Vision, you might send a request like this:
text
import requests
url = "https://api.sarvam.ai/v1/vision/document"
headers = {"API-Key": "YOUR_API_KEY"}
files = {"image": open("pan_card_hindi.jpg", "rb")}
data = {"language": "hi", "extract_structured": "true"}
response = requests.post(url, headers=headers, files=files, data=data)
print(response.json())
# Output: { "text": "Extracted Hindi text...", "fields": {"name": "..."} }
Apps, Chat, and User Experience
sarvam ai app refers to the Indus app, launched in early 2026 for Android and iOS . It is Sarvam’s direct-to-consumer play to rival generic chatbots.
User Experience:
The interface is clean and minimal. Upon signing up (via mobile number or email), users are greeted with a chat prompt. Key features include:
- Voice Input: A prominent microphone icon allows users to “send a voice note” instead of typing, aligning with their voice-first thesis .
- Indian Language Fluency: The app is claimed to be fluent in 11 Indian languages. In practice, this means it can understand and respond in languages like Hindi, Tamil, and Kannada, often mixed with English (Hinglish).
- Internet Search: The sarvam ai chat experience includes the ability to scan the internet for real-time information, moving beyond just static knowledge .
For a user in a Tier-2 city, the experience of asking “Help me cook biryani” in a mix of Hindi and English feels more natural than on ChatGPT, which defaults to English.
Careers, Jobs, Salaries, and Offices
For those searching for sarvam ai jobs or sarvam ai careers, the company represents an exciting opportunity to work on deep-tech problems.
Roles:
Hiring typically focuses on:
- Research: AI/ML researchers with expertise in NLP, speech synthesis, and multilingual models.
- Engineering: Backend engineers, infrastructure engineers (to manage GPU clusters), and full-stack developers.
- Product and Go-to-Market: Product managers and sales executives to help enterprise clients integrate the API.
sarvam ai salary
Salary data for a private startup is not officially disclosed. However, based on market rates for well-funded AI startups in Bengaluru:
- AI Research Scientists with PhDs can expect ₹40-80 Lakhs per annum (plus equity).
- Senior Software Engineers typically range between ₹35-60 Lakhs per annum.
- Internships or entry-level roles are compensated competitively, though one should be aware that a separate entity named SarVm.ai (note the spelling) has posted internships at ₹8,000-15,000 per month, but this is an unrelated e-commerce platform, not the AI startup .
sarvam ai office
The primary sarvam ai office is located in Bengaluru, Karnataka. While they maintain a physical presence for core teams, like many tech startups, they support remote or hybrid work for certain roles.
Public Perception and Backlash
No profile of an ambitious indian ai startup sarvam ai is complete without discussing the sarvam ai backlash of 2025.
The Incident:
In May 2025, Sarvam AI released a model called Sarvam-M. It was benchmarked against global standards. However, critics quickly pointed out that Sarvam-M was a fine-tuned version of Mistral (a French AI company’s model), not a model “trained from scratch” as many had anticipated for a “sovereign AI” .
A prominent investor, Deedy Das, posted on X that the model’s Hugging Face downloads were only 23, calling it “embarrassing” . This sparked a debate about whether Sarvam was overhyped and whether they were actually building foundational tech or just wrapping existing open-source models.
The Response:
Co-founder Pratyush Kumar responded with a single line: “…when we start training the sovereign model, all of this will be put to use. ” .
This was a tacit acknowledgment that Sarvam-M was a stop-gap product. The real test would be the government-backed, from-scratch model. Eight months later, when they launched the 30B and 105B models (trained from scratch) in 2026, the same critic admitted, “I was wrong” .
Takeaway: The backlash highlights the tension between delivering products quickly and the long, expensive grind of foundational research. Sarvam navigated this by managing expectations and eventually delivering the goods.
Share Price and Company Status
A common query is sarvam ai share price. This is a critical point of clarity: Sarvam AI is a privately held company. It is not listed on any public stock exchange like the BSE or NSE.
Therefore, there is no public “share price” that you can look up on a trading app. The valuation, reportedly around $200 million as of early 2026, is an internal figure based on its last funding round .
- Disclaimer: This information is for educational purposes only and does not constitute financial advice. Investment in private companies is typically limited to accredited investors and venture capital firms.
How Good Is Sarvam AI? Our Criteria and Verdict
Is Sarvam AI any good? To answer this, we need a clear methodology.
Methodology Box
How we evaluated: We analysed recent, third-party reported benchmark data (e.g., olmOCR-Bench from February 2026), user reviews of the Indus app, and the technical specifications of their open-source releases. We weighed these against the needs of three user profiles: 1) a Hindi-speaking consumer, 2) an enterprise developer, and 3) a job-seeking AI researcher. We prioritized verified performance metrics over marketing claims.
Criteria:
- Performance: How well does it handle Indian languages?
- Accessibility: Is it easy to use and affordable?
- Innovation: Is it truly novel or just a copy?
- Trust: Is the company stable and transparent?
The Verdict:
- For Consumers: The Indus app is a solid alternative to ChatGPT for Indian languages. It’s free and voice-native, making it more accessible. However, its feature set (e.g., image generation) is currently less extensive than global competitors .
- For Developers/Enterprises: This is where Sarvam shines. Sarvam AI is particularly good if you need to process documents in Indian scripts. Their Vision model’s 93.28% accuracy on OmniDocBench suggests it is best-in-class for this niche . Pricing and latency need to be evaluated on a case-by-case basis, but the free API tier for testing lowers the barrier.
- For India: It is a critical player in building sovereign AI capabilities, reducing dependence on foreign models for sensitive government data.
Pros: Deep focus on Indic languages, strong founding team, government backing, competitive OCR benchmarks.
Cons: Consumer app is still maturing, faces massive competition from well-funded US tech giants, business model dependent on enterprise adoption.
India’s AI Landscape and Identity
Which is India’s own AI? This question has no single answer. “India’s own AI” isn’t a single model but an ecosystem.
- Sarvam AI: Positions itself as the builder of foundational models from scratch within India.
- Krutrim: Backed by Ola’s Bhavish Aggarwal, is another high-profile entrant building its own LLMs.
- AI4Bharat: The non-profit research group that acts as a talent and data nursery for many of these startups .
- Bhashini: The government’s National Language Translation Mission, which provides the digital infrastructure.
Sarvam AI fits into this landscape as a key private player with strong government ties, attempting to commercialize the research originating from academia and public institutions .
Future of Work: Which 3 Jobs Will Survive AI?
A question that often comes up in discussions about Indian AI startups is job security. Which 3 jobs will survive AI? Based on current technological trajectories, we believe these roles are resilient because they require deep human interaction and complex physical dexterity.
- Skilled Tradespeople (Electricians, Plumbers, Carpenters): AI can design a circuit or diagnose a leak, but it cannot crawl under your sink to fix a pipe or rewire a fuse box in a 50-year-old building. These jobs require physical problem-solving in unstructured environments—a challenge robotics is far from mastering.
- Healthcare Professionals (Doctors, Nurses, Caregivers): While AI excels at diagnosis from scans, the role of a doctor or nurse involves empathy, bedside manner, and complex ethical decisions. A patient wants human reassurance when receiving difficult news. The “care” aspect of healthcare is fundamentally human.
- Leadership and Strategy Roles (CEOs, Judges, Policy Makers): These roles involve navigating ambiguity, weighing competing stakeholder interests, and making judgment calls based on values, not just data. AI can provide decision support, but accountability and leadership are human prerogatives.
How to survive? Future-proof your career by focusing on skills AI lacks: empathy, ethical judgment, physical manipulation, and creative strategy.
Getting Started Checklist
Ready to try Sarvam AI? Here is a simple checklist:
- [ ] Step 1: Test the Consumer App. Download the “Indus” app from the Play Store or App Store. Try asking it questions in your local language using voice .
- [ ] Step 2: Explore the API. Visit the Sarvam AI website. Look for the “Developers” or “API” section to see if the free tier is still active.
- [ ] Step 3: Pilot a Use Case. If you’re a business, identify a small project (e.g., digitizing 100 Hindi forms) to test against the Sarvam Vision API.
- [ ] Step 4: Evaluate Pricing. Contact their sales team to get a custom quote based on your volume and compare it to the cost of using a general-purpose OCR like Google Cloud Vision.
- [ ] Step 5: Security Review. If handling sensitive data, review their data handling and compliance policies to ensure they meet your standards for sovereign AI requirements.
FAQs
Is Sarvam AI api free?
As of early 2026, Sarvam AI offered a free trial period for its Document Intelligence API. While ongoing costs may apply for production usage, developers can typically sign up for a free tier with limited credits to test the functionality. For the latest information, it is best to check their official documentation .
Is Sarvam AI any good?
Yes, particularly for Indian language tasks. Independent benchmarks show its Vision OCR model outperforms global competitors like Google Gemini on Indian script recognition. For general-purpose chat, the Indus app is a strong, free alternative for Indian users, though it lags behind ChatGPT in advanced features like image generation .
Who funds Sarvam AI?
Sarvam AI is funded by top-tier venture capital firms. The primary investors are Lightspeed, Peak XV Partners, and Khosla Ventures. It also received subsidized computing power from the Indian government under the IndiaAI Mission in exchange for equity .
Is Sarvam free to use?
The consumer chatbot app “Indus” is free to download and use. For businesses using the API, there may be a free trial tier, but extensive commercial use will likely incur usage-based fees .
What is the legal name of sarvam AI?
The legal name of the company is Sarvam AI Pvt. Ltd. (Private Limited). It is registered in India and operates out of Bengaluru.
Which is India’s own AI?
There is no single “India’s own AI.” It is a collective effort. Sarvam AI is one of the leading contenders, building foundational models. Other notable mentions include Krutrim and the government-backed Bhashini initiative. Sarvam distinguishes itself by building models “from scratch” trained on Indian data .
Which 3 jobs will survive AI?
Jobs that require high levels of human interaction, physical dexterity, and ethical judgment are most likely to survive. These include 1) Skilled Trades (Electricians), 2) Healthcare Professionals (Nurses), and 3) Leadership Roles (Judges/CEOs) .
Conclusion and Call to Action
Sarvam AI represents a bold, necessary step in the global AI landscape. It is not trying to beat OpenAI at its own game; it is trying to build a different game—one that is played in Tamil, Hindi, and Bengali, and accessible via a voice note on a budget smartphone. The journey has had its share of controversy, but the release of their open-source, from-scratch models in 2026 proves they are serious about the long haul.
Whether you are a developer looking to build the next great Indic app, a business needing to process local language documents, or a job seeker wanting to work on meaningful tech, Sarvam AI is a name you need to know.
Ready to build with Indic AI? Sign up for the Sarvam AI API waitlist here (placeholder). Or, to dive deeper into the Indian tech ecosystem, check out our guide on Top Indian DeepTech Startups to Watch in 2026.
You can also check our other blog post
ElevenLabs AI Voice Tools: A Complete Guide for Creators
Nano Banana AI 3D Figurines How to Make – A Beginner’s Guide
Durable AI Website Builder: Features, Pros/Cons, Pricing, and Best Use Cases
Agentic AI: A Practical Guide to Benefits, Use Cases, and Implementation
