OpenAI keeps India API data in India for GPT-6 Luna

OpenAI introduced local inference (the step where a model makes answers) for its API in India. The API calls, and the work to make the answers, stay in India for some workloads. It is first available for the GPT-6 Luna model. India is the third market with this option, after Australia and Japan. India has more than 100 million ChatGPT users each week.

What is new

Customers use the India endpoint, which is the access point of the API. India joins Australia and Japan as markets with API inference residency. Nitin Bawankule, enterprise head of OpenAI India, says that this gives businesses more control of where data is used.

The numbers

The source gives no price, no speed test and no full set of workloads. The one number is more than 100 million users each week. In August 2026, Anthropic announced that it will give in-country inference for Claude to India through Amazon Bedrock.

What is not known

The source does not tell when more models get India residency. It says that Indian startups such as Sarvam AI use local hosting against firms from other countries. This edge can become less strong.

Insight: Location is no longer an edge in India

Sarvam AI used local hosting against firms from other countries. OpenAI and Anthropic offer in-country inference. OpenAI also adds Indian language features, and ads with Razorpay, a payment company. This reads as OpenAI removing the data rule that stopped banks. Price and language then give the edge. If you sell AI in India, location does not set you apart.