The short version
Ranked on shipped product, not funding. Most lists rank by money raised, which tells you who investors backed rather than who built something
1. Sarvam AI - the only one publishing both open weights under Apache 2.0 and per-token pricing. Rs 15 per million input tokens on its 105B
2. Krutrim - the most funded at $524M and India's first AI unicorn, going full-stack with cloud and compute
3. Emergent - launch to unicorn in just over a year, $130M Series C in July 2026
The state programme: four companies hold IndiaAI Mission mandates. BharatGen took the largest allocation at Rs 988.6 crore
The honest caveat: only Sarvam publishes pricing you can check. For the rest, treat figures as reported
There is no shortage of lists of Indian AI companies, and almost all of them rank by funding. That is easy to compile and tells you very little: money raised measures investor conviction, not whether anything works. This list ranks on what a reader can actually use, try or inspect today - published pricing, open weights, a product you can sign into - with funding as context rather than as the score.
| Company | What it ships | Funding / status |
|---|
| 1Sarvam AI | Open weights under Apache 2.0, public API with posted rupee pricing, speech and vision models, a consumer app | $275M. IndiaAI Mission pick |
| 2Krutrim | Indian-language models plus cloud and compute, full-stack | $524M, highest funded. India's first AI unicorn |
| 3Emergent | Builds working software from a description | $230M. Unicorn July 2026 |
| 4Kore.ai | Enterprise conversational and agent platform | $220M |
| 5Gnani.ai | Speech-to-speech model for Indian languages | IndiaAI Mission pick |
| 6BharatGen | Multilingual foundation models, state-backed | Rs 988.6 crore, largest IndiaAI allocation |
| 7Avataar | Video generation model | IndiaAI Mission pick |
| 8Yellow.ai | AI agents for customer and employee support | Application layer, exports across Asia |
| 9Uniphore | Enterprise conversational AI and analytics | Longest-established of the group |
| 10Fractal | AI products and decision intelligence | Enterprise products and services |
1. Sarvam AI
Sarvam is first because it is the only company on this list where you can read the prices, download the weights and try the app without speaking to anyone. Founded in Bengaluru in August 2023 by Vivek Raghavan and Pratyush Kumar, both previously with AI4Bharat at IIT Madras, it raised about $41 million in a combined seed and Series A in December 2023 led by Lightspeed, with Peak XV and Khosla Ventures. Inc42 puts the total at $275 million, second in India.
Sarvam-30B and Sarvam-105B were announced on 18 February 2026 and open-sourced under Apache 2.0 in early March, with weights on Hugging Face and AIKosh. The 105B is a mixture-of-experts model with a 128,000 token context window. A consumer app built on it, Indus, reached beta on iOS, Android and web on 20 February 2026. Alongside the language models it ships Saaras V3 for Indian-language speech and Sarvam Vision for document understanding and Indian-script OCR, and it announced Sarvam Kaze, wearable glasses supporting more than ten Indian languages, for May 2026.
| Service | Price | Notes |
|---|
| Sarvam 105B | Rs 15 in / Rs 5 cached / Rs 60 out per 1M tokens | Open weights, Apache 2.0 |
| Speech to text | Rs 0.49 per minute | Real-time, streaming and batch |
| Text to speech, Bulbul v4 Flash | Rs 1.00 per 1,000 characters | Bulbul v3 at Rs 3.00 |
| Translation | Rs 2.00 per 1,000 characters | Formal and colloquial |
| Document digitisation | Rs 0.50 per page | Max 10 pages per job |
| Dubbing | Rs 40 per minute of source media | Rs 80 with editor flow |
| Free credits | Rs 100 for verified users | No monthly fee on pay as you go |
Why the pricing matters more than the benchmarks: Rs 15 per million input tokens is roughly $0.17. Sarvam also serves third-party models on the same API at posted rates - GLM-5.3 at Rs 126 and Rs 390, Kimi K3 at Rs 270 and Rs 1,350 - which makes it one of the few places you can compare an Indian model against a frontier one on the same bill. Full details on the
Sarvam AI listing.
2. Krutrim
Krutrim is the Ola group's AI company, India's first AI unicorn, and by Inc42's ranking the highest-funded Indian AI startup at $524 million - nearly double Sarvam. The money reflects the strategy: Krutrim is attempting the full stack, models plus cloud plus compute, rather than models alone. That is a far more capital-intensive bet than anything else on this list and the reason its funding requirement is in a different class.
3. Emergent
Emergent builds working software from a natural-language description. Founded by Dunzo co-founder Mukund Jha with Madhav Jha, it raised $70 million at a $300 million valuation in January 2026 from Khosla Ventures and SoftBank Vision Fund 2, tripling its valuation, then took a $130 million Series C in July that made it a unicorn just over a year after launch - India's third AI unicorn of 2026. Inc42 puts the total at $230 million.
The IndiaAI Mission four
The government's IndiaAI Mission backs indigenous foundation models with funding and compute, and four companies have been named publicly for specific model mandates. Sarvam AI for the 30B and 105B language models, Gnani.ai for a speech-to-speech model, BharatGen for multilingual foundation models and Avataar for video generation. As of July 2026 the mission had sanctioned 93.18 lakh GPU hours across 237 research projects.
BharatGen is the one to watch on money. Led from IIT Bombay, it secured Rs 988.6 crore under the mission, reported as the largest single share of MeitY's Rs 1,500 crore allocation. That is public funding at a scale comparable to the venture rounds raised by India's largest private AI companies, directed at models intended as a national resource rather than a commercial asset.
The speech mandate is the one most likely to produce something global models cannot match. Speech recognition trained overwhelmingly on English degrades badly on Indian languages, accents and the code-switching between them that is normal in actual Indian conversation. A speech-to-speech model also avoids the transcribe, translate, re-synthesise chain that compounds errors at every hop.
The application layer, which got there first
Kore.ai at $220 million raised, Yellow.ai, Uniphore and Fractal occupy the second half of this list, and their position reflects the ranking method rather than their size. All four are substantial businesses selling to enterprises, several of them exporting long before the current funding cycle began. What they do not do is publish prices or models, so there is less for a reader to verify or try.
They matter to the overall picture for one reason: India's AI sector was commercially viable at the application layer years before the sovereign model programme existed. The foundation-model story is newer and louder, but it is not the whole industry.
How we ranked this
Published pricing counts most, because a company that posts rates is a company you can evaluate without a sales process. Open weights count next, particularly under a permissive licence. A product you can sign into and use counts after that. Funding is context only - it appears in the table because readers expect it, not because it determines position.
What we verified and what we did not. Sarvam's pricing and model details come from its own documentation and public records. Funding figures come from Inc42's ranking and TechCrunch's reporting on Emergent. The IndiaAI Mission mandates and the GPU-hour figure come from reporting dated 23 July 2026, and BharatGen's Rs 988.6 crore allocation from reporting on the MeitY disbursement. For the remaining companies we did not verify pricing against vendor pages, and their listings say so rather than carrying a figure we could not check.
FAQ
Which Indian AI company is the biggest?
By funding, Krutrim: Inc42 ranks it first among Indian AI startups at $524 million raised, and it was India's first AI unicorn. By public money, BharatGen, which took Rs 988.6 crore under the IndiaAI Mission. By what you can use without a sales call, Sarvam AI, which publishes open weights and per-token pricing.
What is the IndiaAI Mission?
A government programme backing indigenous foundation models with funding and compute. Four companies have been named publicly: Sarvam AI for 30B and 105B language models, Gnani.ai for speech-to-speech, BharatGen for multilingual foundation models and Avataar for video generation. BharatGen took the largest single share at Rs 988.6 crore of MeitY's Rs 1,500 crore allocation. As of July 2026 the mission had sanctioned 93.18 lakh GPU hours across 237 research projects.
Are Indian AI models actually cheaper?
Sarvam's published rates are the only ones on this list you can check directly, and they are low: Rs 15 per million input tokens and Rs 60 per million output for its 105B model, roughly $0.17 and $0.68 at current rates. It also resells third-party models on the same API, where GLM-5.3 runs Rs 126 and Rs 390 and Kimi K3 runs Rs 270 and Rs 1,350.
Can I download and self-host an Indian model?
Sarvam-30B and Sarvam-105B were open-sourced under Apache 2.0 in early March 2026 with weights on Hugging Face and AIKosh, so yes. Sarvam-M from May 2025 is also Apache 2.0. Apache 2.0 permits commercial use, which not every open-weight release does.
How many Indian AI unicorns are there?
Krutrim was the first. Emergent became the third of 2026 with its $130 million Series C in July, just over a year after launch, having tripled its valuation to $300 million in January.
Why are enterprise companies ranked below startups here?
Because the ranking is on shipped, inspectable product. Kore.ai, Yellow.ai, Uniphore and Fractal are substantial businesses with real customers, but their pricing is quoted per deployment and their models are not published, so there is less a reader can verify or try. That is a ranking choice, not a verdict on the companies.
Sources
Sarvam AI API pricing ·
Sarvam AI company record ·
Inc42, top funded Indian AI startups ·
TechCrunch on Emergent's unicorn round, 15 Jul 2026 ·
BharatGen ·
Reporting on the IndiaAI Mission selections, 23 Jul 2026
More on the programme behind four of these companies: what the IndiaAI Mission has funded, by the numbers.