Image: Wikimedia Commons / FoxTPNL, CC BY 4.0 — the OpenAI logo. The rollout covers the OpenAI API, not the ChatGPT app.
TL;DR: OpenAI API local processing in India explained: from 12 November 2026, eligible businesses can run supported GPT-6 Luna API workloads entirely inside India — request, model computation and storage — instead of only storing data here. It needs OpenAI-approved Zero Data Retention, costs about 10% more than standard rates, and does not cover ChatGPT or other models yet.
Key Highlights
- OpenAI announced on Thursday, 12 November 2026 that its API now supports local processing and inference in India — the model computation itself happens in the country, not just data storage.
- At launch the only supported model is GPT-6 Luna, through the Responses API and Chat Completions API, via the India regional endpoint in.api.openai.com.
- Access is gated: organisations need active Zero Data Retention (ZDR), approved by OpenAI. Modified Abuse Monitoring on its own is not enough.
- Regional processing carries a 10% price uplift over GPT-6 Luna's list rates — roughly $0.11 per million input tokens and $0.55 per million output tokens, on standard pay-as-you-go only.
- India joins Australia and Japan as the latest additions to OpenAI's API inference residency; the US, Europe and the UAE already had it. ChatGPT is not covered.
OpenAI API local processing in India explained: what changed on 12 November
OpenAI has switched on local processing for its application programming interface (API) in India. For eligible businesses and developers, supported API requests — and the model computation that generates the response — now take place on infrastructure in India, and supported customer content is stored in India too.
That is a real change from what existed before. Since May 2025, OpenAI has offered data residency in India for ChatGPT Enterprise, ChatGPT Edu and the API platform — but that covered where customer content is stored. Where the model actually read that content and produced an answer was a different question, and the answer was usually outside India. This announcement closes that gap for one model: the request goes in, the model runs, and the answer comes out, all within the country — for the workloads the rollout covers.
The announcement, reported on 12 November by PTI, The Hindu BusinessLine, Business Standard, Inc42 and others, is aimed squarely at enterprises rather than individual ChatGPT users. OpenAI's head of enterprise for India, Nitin Bawankule, said local inference "gives them more control over where supported workloads run and helps address a practical barrier to adoption" — the barrier being that many regulated Indian organisations could not move an AI project from pilot to production while the processing happened abroad.
What can run in India: GPT-6 Luna only, for now
The scope at launch is deliberately narrow. Everything in the table below is what OpenAI's documentation lists for India today; anything outside it still runs on the global service.
| What | Covered in India at launch |
|---|---|
| Model | GPT-6 Luna only (model ID gpt-6-luna) |
| APIs | Responses API and Chat Completions API |
| Inputs / outputs | Text and image inputs; text output |
| Endpoint | in.api.openai.com — via a new regional API project with India selected for storage and processing |
| Processing location | Supported API request intake and model computation take place in India |
| Storage | Supported customer content stored in India |
| Not covered | ChatGPT, other models, OpenAI-hosted tools, Batch and Flex processing |
GPT-6 Luna is the cheapest model in OpenAI's current GPT-6 family — it was released on 22 September 2026 with a 1.05-million-token context window, aimed at focused, high-volume tasks such as summarising documents, analysing scanned material and drafting responses for staff to review. That profile fits the workloads the rollout is designed for: document-heavy enterprise jobs where the content is sensitive but the task itself is routine.
One thing the announcement does not say is where in India the processing runs. OpenAI has not identified the facility serving these requests, and the rollout is not linked to the domestic data-centre capacity OpenAI and the Tata Group said in February 2026 they planned to develop under the "OpenAI for India" initiative. Until OpenAI says otherwise, treat this as a routing and residency commitment, not a data-centre launch.
Illustrative image: a data centre. Image: Wikimedia Commons / Fleshas, CC BY-SA 3.0. OpenAI has not identified the facility serving its India API requests.
Who can use it: eligibility and setup
This is not a toggle every developer account gets. To use regional processing in India, an organisation must:
- Have active Zero Data Retention (ZDR) — OpenAI's control under which it does not retain API inputs and outputs. ZDR is subject to OpenAI approval, and for non-US regions OpenAI also requires approval for abuse-monitoring controls and a Modified Retention amendment. Modified Abuse Monitoring alone is not sufficient for inference in India.
- Create a new API project in the OpenAI platform, selecting India as the location for both data storage and data processing.
- Route supported requests through the India regional endpoint (in.api.openai.com) using the Responses or Chat Completions API with GPT-6 Luna.
There is also a limit worth understanding before planning a migration. OpenAI distinguishes customer content from system data: the residency controls apply to in-scope customer content, while system data such as billing information, usage statistics and certain metadata may be processed or stored outside the selected region. And ZDR itself disables the option to save responses through the two supported endpoints, subject to documented limitations. So a bank or insurer can now answer "where does the model process this request?" with "in India" — but the compliance assessment still has to cover the whole workflow: where the application retrieves documents, what its logs capture, where it saves generated outputs, and whether any other vendor touches that data.
What it costs in India
India supports only standard pay-as-you-go processing for this rollout — discounted Batch and Flex processing are not available through the India region. And regional endpoints carry a price uplift: OpenAI's documentation puts it at 10% for eligible models released on or after 5 March 2026. GPT-6 Luna, released in September 2026, qualifies, so the effective India rates work out as follows.
| GPT-6 Luna, per 1M tokens (Standard) | Global list rate | India regional (list + 10%) |
|---|---|---|
| Input (prompts up to 272K tokens) | $0.10 | ~$0.11 |
| Output | $0.50 | ~$0.55 |
| Cached input | $0.01 | ~$0.011 |
These are calculated from OpenAI's published list rates plus the documented 10% uplift — OpenAI has not published a separate India price card — so re-check the pricing page before budgeting a production workload. Above 272,000 input tokens the whole request moves to higher long-context rates ($0.20 input / $0.75 output globally), and the uplift applies on top of those too. At these volumes the money is small in absolute terms — a million input tokens costs about eleven cents — but the premium matters at production scale, and teams should model costs at the regional rate rather than assuming they can use discounted processing modes in India.
For comparison, there is already a second route to in-country OpenAI inference in India: Amazon Bedrock has offered the GPT-5.6 Terra and Luna models with India geographic cross-Region inference since around September 2026, routing requests only between its Mumbai and Hyderabad regions. That is a different generation of models (GPT-5.6, not GPT-6) on a different platform, but it is worth knowing about if an organisation already runs on AWS.
Why it matters for India readers
The timing is not accidental. India is, in Inc42's words, OpenAI's largest market for ChatGPT, and the company has been building its India presence all year — the February 2026 "OpenAI for India" initiative covers infrastructure, enterprise adoption, skills and local partnerships, and OpenAI has been opening offices in Mumbai and Bengaluru. Local processing removes one of the last structural objections Indian enterprises had to deploying OpenAI's API in production.
The sharpest effect is in financial services. The Reserve Bank of India has mandated local processing and storage of payment data, and banks, insurers and fintech firms have long had to keep sensitive customer data inside India. Until now, an organisation in that position could store data in India but could not honestly say the AI processing happened here — so the AI project stayed a pilot. With inference residency, a document-review or claims-processing workflow built on GPT-6 Luna can clear that bar, provided the rest of the workflow stays in-country too.
The legal backdrop is the Digital Personal Data Protection Act, 2023, which introduced data-localisation requirements for certain categories of personal data. Local inference does not by itself make an organisation DPDP-compliant — notices, consent, breach reporting and the rest of the Act still apply — but it removes the processing-location question from the list of things a compliance team has to fight about. The same logic applies to healthcare and public-sector workloads, where data is often not allowed to leave the country at all.
It is also worth keeping the scale of the change in proportion. This is one model, two APIs, and a gated enterprise feature — not a general "ChatGPT now runs in India" moment. The everyday ChatGPT user notices nothing. What changes is that a specific class of Indian business — regulated, data-sensitive, already building on the API — can now move from experimenting with AI to running it in production without sending its data abroad to be processed.
Should you care?
✅ Yes, if you build on the OpenAI API in India — check whether your organisation qualifies for Zero Data Retention, test GPT-6 Luna against your workload through the India endpoint, and cost it at the regional rate. ⏳ Wait, if you use ChatGPT or another model — nothing changes for you yet, and there is no announced date for wider model coverage.
If you are a developer, the practical next step is to open the OpenAI platform, check your organisation's ZDR status, and confirm GPT-6 Luna is listed for the India endpoint in the data-controls guide before planning around it. If you are a business user, the question to ask your vendor or IT team is simpler: does our AI workflow still send data abroad to be processed, and does this announcement change the answer?
FAQs
What is OpenAI API local processing in India?
It is OpenAI's API inference residency for India, announced on 12 November 2026. For eligible API projects, supported requests and the model computation that generates the response take place in India, and supported customer content is stored in India — extending the earlier India data residency, which covered storage only.
Which OpenAI model supports local processing in India?
Only GPT-6 Luna at launch, through the Responses API and Chat Completions APIs, with text and image inputs and text output. No other model, and no ChatGPT access, is covered yet, and OpenAI has not announced when more models will be added.
How much does OpenAI API local processing in India cost?
India supports standard pay-as-you-go pricing only, with a 10% regional uplift on GPT-6 Luna's list rates. That works out to roughly $0.11 per million input tokens and $0.55 per million output tokens for standard prompts, calculated from the published list rates — re-check OpenAI's pricing page before budgeting.
Does OpenAI local processing in India apply to ChatGPT?
No. The rollout covers direct use of the OpenAI API only. ChatGPT, ChatGPT Enterprise and ChatGPT Edu are not part of this announcement, and OpenAI-hosted tools are unavailable for regional processing in India.
Takeaway
OpenAI's India API now keeps both the data and the thinking at home — for one model, for eligible enterprises, at a 10% premium. For regulated Indian businesses that was the missing piece; for everyone else, it is a sign of where OpenAI's India bet is heading.
Curious how the other labs are localising for India? Read our explainer on Anthropic's rival India-priced office suite — and if you want the wider OpenAI picture, see OpenAI's push beyond chat into travel booking and OpenAI's 35-second kill switch for runaway research models.
Sources
PTI via Rediff — OpenAI's Local AI Processing in India: What It Means for Data Localisation (12 Nov 2026)
The Hindu BusinessLine — OpenAI announces local AI processing in India with API inference residency (12 Nov 2026)
Business Standard — OpenAI to process API data, run AI inference locally for Indian companies (12 Nov 2026)
Inc42 — OpenAI Brings API Inference Residency to India to Deepen Enterprise AI Adoption (12 Nov 2026)
DQIndia — OpenAI brings GPT-6 Luna API processing to India (12 Nov 2026)
FoneArena — OpenAI API now processes and runs locally in India (12 Nov 2026)
OpenAI — API pricing (GPT-6 Luna list rates)
AWS — Introducing OpenAI models on Amazon Bedrock for in-country inferencing in India
Comments
Post a Comment