GPT-5.2 leads the Indic LLM Arena with 1,314, ahead of GPT-5.1 (1,298) and Gemini 3 Flash (1,288)

What the chart shows

The Indic LLM Arena, led by AI4Bharat at IIT Madras, scores models through crowd-sourced evaluation in more than 20 Indian languages, covering language quality, cultural grounding and safety. GPT-5.2 scored highest of the 27 models at 1,314, followed by GPT-5.1 (1,298) and Gemini 3 Flash (1,288).

OpenAI and Google models hold the top 11 places. The highest-scoring open-weight model, Qwen3-Next-80B, scored 1,156, and Llama-4-Maverick 1,108. The lowest score was 974, for Qwen3-235B-A22B-Thinking.

Key numbers

How to read it

• Scores are relative ratings from the arena’s crowd-sourced evaluation, so the differences between models matter more than the absolute numbers.
• Open-weight models in the list are the Qwen, Llama, Gemma and Granite models; the GPT and Gemini models are proprietary.
• Results apply to the 27 models in the AI Index’s snapshot of the leaderboard.
• Arena ratings have no true zero, so the chart shows dots on an axis starting at 950: compare the differences between models.