OpenAI Introduces IndQA Benchmark for Indian Languages and Culture

OpenAI has introduced IndQA, a new benchmark designed to measure how well AI systems understand Indian languages and culture, according to an announcement on the company’s website. The benchmark aims to evaluate reasoning and contextual understanding across a broad range of cultural topics rather than focusing solely on translation or multiple-choice tasks.

IndQA includes 2,278 expert-authored questions in 12 Indian languages, covering 10 cultural domains such as literature, food, history, law, and religion. The dataset was created with contributions from 261 Indian experts, including linguists, journalists, artists, and scholars. Each question includes a rubric for grading responses, an ideal answer, and an English translation for consistency.

Questions were tested against several of OpenAI’s advanced models, including GPT‑4o, OpenAI o3, GPT‑4.5, and GPT‑5, to ensure difficulty and room for improvement. Evaluation results show that the GPT‑5 Thinking High model achieved the highest IndQA score among current models, though performance varied across languages and domains.

OpenAI stated that IndQA will be used to track progress in multilingual and culturally grounded reasoning and may serve as a model for similar benchmarks in other regions and languages.