Krutrim
India's first generative-AI unicorn, now a rupee-billed GPU cloud with open-weight Indic models
Overview
Krutrim (Sanskrit for "artificial") is the AI company Ola's Bhavish Aggarwal launched in April 2023 with Krishnamurthy Venugopala Tenneti, headquartered in Bengaluru. It calls itself India's first AI unicorn; Fractal Analytics reached that mark in 2022, so the safer title is India's first generative-AI unicorn. The company built Krutrim-2, a 12-billion-parameter multilingual LLM trained on English and 22 Indian languages using a Mistral-NeMo architecture with a 128K-token context window, and released open-weight Indic models covering text, translation (Krutrim-Translate), vision (Chitrarth), OCR (Chitrapathak-2) and speech translation (Dhwani) under its own community licence.
After reaching a $1 billion valuation in January 2024, Krutrim realigned from late 2025: it paused chip-design work, media reports say it halted its next foundation model, and it wound down its consumer assistant Kruti (launched June 2025, offline since April 2026). In May 2026 it repositioned as an AI cloud platform offering GPU pods, dedicated GPUs, CPU VMs and inference infrastructure from Bangalore and Hyderabad, billed in rupees with Indian data residency. It reported about ₹3 billion (roughly US$31.5M) in FY2026 revenue and its first annual net profit, both company figures, while its workforce contracted from 550+ to around 150.
Pricing
Pricing shown for reference only. These figures reflect RECATOOLS research as of 28 Sep 2026 and may be out of date or incomplete. This is not financial or purchasing advice — always confirm the current price on the provider’s official website before making any decision.
- 1–32 vCPU configs
- Down to ₹2.10/hr on 1-yr reserve
- Block + object storage add-ons
- Tiny/Nano/Mini to 8x A100
- ₹170/hr for a full A100
- Indian data residency
- Tiny to 8x H100
- ₹213/hr for a full H100
- A100 and H100 x1–x4 nodes
- H100 x1 ₹213/hr, ₹173 on a 1-yr term
- Custom enterprise quotes
Use cases
What you can produce with Krutrim
- A multilingual chatbot in Hindi, Tamil, Bengali or 19 other Indian languages built on the open-weight Krutrim-2 model
- A fine-tuned Indic LLM deployed on Krutrim Cloud GPU pods with Indian data residency
- English-to-Indic or Indic-to-English translation with the open-weight Krutrim-Translate model, which Krutrim says runs four times faster than distilled IndicTrans2
- Image-and-text understanding in English and ten Indic languages via the Chitrarth vision-language model
- Rupee-billed GPU compute in Bangalore or Hyderabad, from ₹24 an hour for a fractional A100 pod
- Speech-to-text translation between Indic languages and English with the gated Dhwani model
- Indic document OCR with the open-weight Chitrapathak-2 model, released March 2026
ASEAN Perspective
Krutrim in Southeast Asia
Krutrim has no confirmed data centres or commercial partnerships in ASEAN and its product roadmap is explicitly India-first, limiting direct relevance for Southeast Asian users today. That said, its approach — sovereign AI cloud infrastructure, Indic-language model development under a community licence, and GPU compute positioned far below Western hyperscaler pricing — is a template that ASEAN governments and startups are watching closely as they pursue their own digital-sovereignty agendas. India and ASEAN share a large overlap of multilingual, price-sensitive AI consumers underserved by Western LLMs, and any future Krutrim expansion eastward (Singapore, Indonesia) would find a receptive market. For now, ASEAN users should treat Krutrim as a monitor-and-wait: promising infrastructure player, not yet a deployed regional product.
Krutrim became India's first generative-AI unicorn, and its Krutrim-2 model delivers real multilingual capability across 22 Indian languages. For developers building India-facing products, Indian data residency and rupee-billed GPU compute — fractional A100 pods from ₹24 an hour — make a coherent stack, and the open-weight releases (Krutrim-2, Krutrim-Translate, Chitrarth, Dhwani, Chitrapathak-2) are genuine research contributions. The pivot story tempers the enthusiasm. The consumer Kruti assistant went offline within a year of launch, chip design is paused, and the company contracted from 550 to around 150 employees. Revenue is now enterprise cloud, and Krutrim has not said how much comes from Ola's own businesses; there is no free tier, and review counts are too thin to judge service quality. For ASEAN or global users seeking a general-purpose AI assistant there is nothing here — this is an India-market infrastructure bet.
What we found
Kruti is gone. Ola Krutrim's consumer AI assistant, which launched in June 2025 with ambitions to be India's answer to a general-purpose chat app, quietly stopped working across the App Store, Google Play and the web in April 2026 — SensorTower put its lifetime downloads under 500,000, a rounding error next to the hype that surrounded its debut.
The retreat runs deeper than one app. Krutrim shed roughly 200 jobs through 2025, shrinking from over 550 employees to around 150–160 by March 2026, and several senior linguistics and leadership hires walked out the door. Its chip-design programme, once pitched as India's answer to Nvidia dependence, is paused by the company's own account, and media reports say work on a Krutrim 3 foundation model has stalled. What's left is a pivot toward Krutrim Cloud — GPU-as-a-service and inference infrastructure billed in rupees, positioned against AWS and Azure on price, though the "up to 25x cheaper" figure long repeated about it appears nowhere on Krutrim's current site or in its pivot announcement.
The technical output that remains is genuinely solid: Krutrim-2, a 12B model on the Mistral-NeMo architecture, still posts competitive results on Indic-language benchmarks, and new open-weight releases followed in March 2026, among them Chitrapathak-2, an OCR model for ten Indic languages plus English. Reviews are too thin to lean on: Techjockey shows Krutrim Cloud at 3.8 from four ratings. But this is now an infrastructure company wearing an AI-unicorn's old branding — judge it as an India-region compute vendor with useful open-weight side projects, not as a challenger to ChatGPT or Gemini.
The cloud runs in two regions, Bangalore and Hyderabad, bills in rupees with itemised invoices, and offers Python, Go, Java and Rust SDKs, AWS-compatible APIs and a Terraform provider. Krutrim reported about ₹3 billion in FY2026 revenue and its first annual net profit, without saying how much came from Ola's own businesses. There is no free compute tier.
Compiled by RECATOOLS from public sources: vendor documentation and pricing, code repositories and published reporting.
Notable facts
- Krutrim is a Sanskrit word meaning 'artificial' — making it one of the few AI companies whose name is itself a translation of 'AI'.
- It became India's fastest-ever unicorn, reaching a $1 billion valuation within months of its first external funding round in January 2024.
- Founder Bhavish Aggarwal personally committed $230 million from his family office toward Krutrim's $1.15 billion funding target — one of the largest founder-led bets in Indian tech history.
- Krutrim-2 was built on Mistral-NeMo architecture and trained on what the company calls the largest known Indic dataset, covering languages spoken by over a billion people who are represented by just 1% of Common Crawl web data.
Frequently asked questions
About this listing
This entry was compiled from publicly available data including Krutrim's official website, press releases, documentation, and reputable third-party publications. RECATOOLS is not affiliated with Krutrim unless explicitly stated.
Third-party AI tools update their pricing, features, availability, and policies frequently. Information here may be outdated by the time you read this — we make reasonable efforts to keep listings current, but cannot guarantee absolute accuracy.
For the latest details, please refer to Krutrim directly →
Spotted something out of date? Suggest an update →
More in Code & Dev Tools
