Indic LLM Arena: A Benchmark for India’s Multilingual AI Future
IIT Madras–supported AI4Bharat has introduced the Indic LLM Arena, a crowd-powered platform designed to assess large language models built for Indian users. This leaderboard establishes a benchmark for how AI systems understand, respond, and behave across India’s vast linguistic and cultural diversity.
Most existing global leaderboards focus primarily on English, overlooking how models perform in Indian languages or in mixed-language conversations like Hinglish or Tanglish. The Indic LLM Arena bridges this gap by evaluating AI models based on three core aspects—language, context, and safety.
It checks whether a model can grasp how Indians naturally speak and switch between languages, whether its responses fit local social and cultural situations, and whether it aligns with India’s values of fairness and sensitivity.
🚀 Announcing the Indic LLM-Arena 🇮🇳
— AI4Bharat (@ai4bharat) November 10, 2025
At AI4Bharat (IIT Madras), our mission has always been clear – build open, inclusive, and world-class AI for Indian languages.
To further this goal, today, we’re introducing the Indic LLM-Arena, a crowd-sourced, human-in-the-loop leaderboard…
This launch aligns with India’s growing efforts toward building sovereign AI capabilities under national initiatives. AI4Bharat envisions the leaderboard as a trusted measure to evaluate both domestic and international language models for their suitability in Indian contexts.
A Public Utility for India’s AI Ecosystem
The platform relies on a human-in-the-loop system. Users can type, speak, or transliterate prompts in any Indian language, view replies from two anonymous AI models, and choose the better response. Thousands of such votes will create statistically reliable rankings, identifying which models best serve Indian users.
AI4Bharat emphasizes that the Arena is more than just a leaderboard—it’s a public resource for India’s AI landscape. Developers can fine-tune Indic models, enterprises can identify the most effective AI tools, and citizens can play a role in shaping what responsible and culturally aware AI should look like for the country.
The team also plans to broaden the platform to evaluate multimodal models that handle text, images, and audio, as well as advanced agent tasks such as search, reading, and tool use—all while keeping the initiative open-source. Supported initially by Google Cloud, the platform is available at arena.ai4bharat.org. AI researcher Adithya S. K., founder of CognitiveLabs, praised the effort, noting that it offered the best Kannada typing experience and reflected the kind of innovation India’s labs should pursue more widely.
Also Read:
Lovable nears 8 million users as the year-old AI coding startup eyes more corporate employees
