A Mixture-of-Experts text-generation model with a strong India-language focus. Model-card-reported benchmark results are not promoted as independent rankings.
verifiedmodelindiclocal
🇮🇳 Built in India evidence hub
Indian & Indic AI
India-first discovery, without sovereignty theatre.
A source-first map of Indian and Indic model work. India relevance is recorded from primary publisher evidence; it is not automatically converted into a DPDP, residency, security or sovereignty claim.
5 verified India-relevant model cards3 primary publishers representedPublisher benchmarks stay publisher-reported
Verified model identities
Indian and Indic model cards
These records pass the same primary-source identity gate as the rest of the public graph.
Large Sarvam model with a first-party model card and downloadable artifacts; publisher benchmarks remain labeled as publisher-reported.
verifiedmodelindiclocal
Base model focused on Hindi, English and Hinglish; the source explicitly notes it is a base model rather than a ready-made chat assistant.
verifiedmodelindichindi
Hindi instruction-tuned model built from OpenHathi and the IndicInstruct dataset.
verifiedmodelindichindi
Instruction-tuned 12B model built on Mistral-NeMo architecture with an India/Indic focus.
verifiedmodelindiclocal
How to read this hub
India relevance is a fact field, not a badge.
◎
Publisher identity
Developer and model identity must resolve to a primary model card, repository or official publisher page.
▥
Benchmarks
Publisher-reported benchmarks are labeled as such and are not converted into independent rankings.
◇
Compliance
No DPDP, data-residency or sovereignty status is inferred from a model being Indian, Indic or locally runnable.
Local inference
Reproducible before impressive.
The India local-LLM benchmark route documents methodology and evidence requirements rather than publishing unsupported hardware or performance numbers.
Policy & governance
Government initiatives and policy
National missions, infrastructure investment, and the regulatory framework AI deployments in India operate under.
IndiaAI Mission (₹10,000 crore)
The IndiaAI Mission, announced in 2024, allocates ₹10,000 crore (~$1.25 billion) to accelerate domestic AI development.
- AIRAWAT — high-performance computing infrastructure for startups, researchers, and academia.
- IndiaAI datasets — curated, multilingual datasets for Indian languages.
- Safe AI — guidelines for responsible AI deployment under the DPDP Act.
- Application development grants for AI applications in healthcare, agriculture, and education.
BHASHINI — India's language translation platform
BHASHINI builds language translation capabilities across 22 Indian languages and 11 global languages, with open APIs for integrating multilingual capabilities into applications.
- 22 Indian languages supported with voice and text translation.
- Open-source models and datasets available via the IndiaAI portal.
- API-first architecture.
DPDP Act and AI deployment
The Digital Personal Data Protection Act (DPDP Act), 2023 governs how personal data — including voice data, classified as biometric data — can be collected, processed, and stored by AI systems in India.
- Explicit informed consent required for voice/biometric data processing.
- Data fiduciary obligations for data security and breach notification.
- Preference for data localization; cross-border data transfers subject to approval.
Definitions
What "India-relevant" actually means here
Three distinct facts get collapsed into one badge on lower-rigor sites. This hub keeps them separate.
◉
Indian-developed
The publisher or lab is headquartered or primarily operating in India — a fact about the organization, not the model's language capability or deployment location.
◈
Indic-capable
The model has documented training or evaluation on one or more Indian languages — a fact about the model's scope, independent of who built it or where it runs.
◆
India-deployable
Inference can run on infrastructure located in India, or the vendor offers an India region — a fact about deployment options, not a compliance guarantee on its own.
A model can be Indian-developed without being Indic-capable (trained primarily on English data), Indic-capable without being Indian-developed (a foreign lab that added Hindi training data), or India-deployable without either (a global model offered through an India-region cloud endpoint). Treating these as one badge — as many product pages do — hides which specific fact is actually true, which is precisely the kind of collapse this hub's evidence gate is designed to prevent.
Resources
Developer resources for India
Datasets, APIs, tutorials, and community resources for Indian AI developers.
Datasets
Indic language datasets, translated corpora, and region-specific data via the IndiaAI portal and Bhashini API.
indiaai.gov.in ↗APIs
Bhashini translation API, and Sarvam and Krutrim models with INR pricing and Hindi documentation.
bhashini.gov.in ↗Communities
Indian AI communities: Hugging Face India, Indic NLP research groups, and local MCP developer forums.
Explore agents ↗