HFAI4Bharat Airavata
Made in IndiaOpen-source Hindi instruction-tuned LLM from IIT Madras, built on Llama architecture with high-quality Indic instruction data.
Best for
- •Researchers and developers needing an open Hindi LLM for fine-tuning.
- •Budget-conscious teams that can self-host for Hindi-specific tasks.
Limitations
- •Smaller parameter count limits complex reasoning.
- •Primarily Hindi — other Indic languages not deeply supported.
- •No tool calling or structured output support.
Use carefully when
- •You need production-grade multilingual or multimodal capabilities.
Quickstart
- Download model weights from Hugging Face (ai4bharat/Airavata).
- Load with transformers library and set Hindi system prompt.
- Fine-tune on your domain data for best Hindi task performance.
Setup checklist
- • API key required: No
- • SDK quality: medium
- • Self-host difficulty: medium
Usage Notes
- • Validate model behavior on your own benchmark slices before rollout.
- • Pin version/provider routes for reproducible outputs.
- • Add logging + fallback routes for high-volume workloads.
Pricing (EUR)
Input / 1M
0,00 €
Output / 1M
0,00 €
Monthly
-
Capabilities
- contextTokens4096
- indicLanguages1
- openWeightYes
- instructionTunedYes
Benchmarks
indic N L U
71.3
hindi M T Bench
68.5
instruction Following
72.1
Community reviews
0 reviews • avg —
No reviews yet.
Samples
textHindi instruction
भारतीय अर्थव्यवस्था पर 200 शब्दों का सारांश लिखें।
Compliance
- License: cc-by-4.0
- Commercial use: allowed
Provenance
- Last verified: 3/3/2026
- Source: https://ai4bharat.iitm.ac.in