OpenAI announced on Thursday, November 12, 2026, API inference residency in India for GPT-6 Luna, one day after bringing the same capability to Japan. With the update, businesses and developers in both countries can process artificial intelligence requests without data leaving national territory.
In practice, both request processing and the model computation that generates the response now happen inside the chosen country. Until now, OpenAI's data residency mainly covered storage: since May 2025, eligible customers could already keep customer content at rest in Japan, India, Singapore and South Korea, but inference — model execution on GPUs — still ran outside the region.
What changes in India
The feature launches initially only for GPT-6 Luna, through the Responses and Chat Completions APIs. To use it, customers create a new API project with India selected for both storage and processing, and then use the country's regional endpoint.
The move targets sectors with data-localization requirements, such as financial services. "Some Indian businesses have not fully deployed AI for important work because they need the information involved to be processed within the country. Local inference gives them more control over where supported workloads run and helps address a practical barrier to adoption," said Nitin Bawankule, Head of Enterprise – India at OpenAI.
The announcement is part of the OpenAI for India initiative, which covers infrastructure, enterprise adoption, skills and local partnerships — including a partnership with Tata Group to develop domestic AI data-center capacity.
In Japan, local inference requires zero data retention
In Japan, the update was announced on November 11 by Takehiro Ogino of OpenAI Japan, in a post on his personal account. GPT-6 Luna is, so far, the only model with local inference in the country, accessed via the regional endpoint jp.api.openai.com.
There are important conditions: using inference in Japan requires OpenAI's prior approval for Zero Data Retention (ZDR), applies only to the Standard pay-as-you-go tier, and does not cover OpenAI-hosted tools such as file search and code interpreter. Under the official specifications, there is also a 10% surcharge on regional endpoint usage for data-residency-supported models released on or after March 5, 2026.
Context
The expansion follows the same logic as the data residency announced in May 2025: separating where content is stored from where it is processed. OpenAI had already brought the feature to Australia and now to Japan and India. The company says it plans to expand supported regions over time.
One caveat is worth noting: local inference does not mean an application's entire processing stays inside the country. System data — such as account, usage and billing information — is not covered by the regional designation, and features outside ZDR's scope keep their own retention rules.
Related: Geekiki already covered GPT-6 Luna in October, when OpenAI made the Decisions API generally available — read Decisions API OpenAI com GPT-6 Luna: decisões até 10x mais rápidas para apps e agentes.
Image credit: OpenAI symbol logo (blossom, in use since February 2025), by OpenAI, via Wikimedia Commons.
Sources
- BusinessFortnight — OpenAI launches API inference residency for GPT-6 Luna in India
- XenoSpectrum — OpenAI Brings GPT-6 Luna Inference to Japan, With Zero Data Retention Required
- OpenAI — Data residency and inference residency for ChatGPT
- OpenAI — Introducing data residency in Asia
Transparency: this article was produced with the help of artificial intelligence, based on cross-checking public posts and the internet sources cited above. We recommend consulting the original sources to confirm the details.
By Muse
