--- title: "OpenAI’s API now runs inside India: one model, at a 10% premium" summary: "What OpenAI’s data-controls guide said on 12 November: in-India processing at in.api.openai.com for GPT-6 Luna only, with Zero Data Retention required and a 10% premium." url: https://www.thetechbag.com/blog/openai-api-inference-india-residency-10-percent format: Note desk: "AI in the enterprise" region: India published: 2026-11-12 facts_checked: 2026-11-12 author: TechBag (The TechBag Journal) tags: ["OpenAI India data residency","OpenAI API India","AI data localisation India","in-country inference","OpenAI in-country inference","RBI data localisation","OpenAI","GPT-6 Luna","in.api.openai.com","data residency","inference residency","Zero Data Retention","Modified Abuse Monitoring","RBI payment data storage","data localisation","Responses API","Chat Completions","Amazon Bedrock","Anthropic","Claude","Australia","Japan","conflict of interest"] cover: https://www.thetechbag.com/blog/covers/openai-api-inference-india-residency-10-percent.png --- # OpenAI’s API now runs inside India: one model, at a 10% premium *What OpenAI’s data-controls guide said on 12 November: in-India processing at in.api.openai.com for GPT-6 Luna only, with Zero Data Retention required and a 10% premium.* Note · AI in the enterprise · By TechBag · Published 2026-11-12 · Facts checked 2026-11-12 · Canonical: https://www.thetechbag.com/blog/openai-api-inference-india-residency-10-percent OpenAI now runs API requests inside India: a project set to India and called at `in.api.openai.com` keeps both the model’s processing and stored customer content in the country, its data-controls guide says. The terms are narrow. One model, GPT-6 Luna; text and image in, text out, through the Responses and Chat Completions APIs only; no hosted tools; Standard pay-as-you-go only; and active Zero Data Retention, where Modified Abuse Monitoring alone is “insufficient”. Regional processing adds a 10% premium.[1][2] Inc42 reported the launch on 12 November, quoting Nitin Bawankule, who heads OpenAI’s enterprise business in India: “Some Indian businesses have not fully deployed AI for important work because they need the information involved to be processed within the country.”[3] Note A disclosure: this piece was researched and drafted with the help of Claude, an AI model made by Anthropic, which is named in this piece. TechBag’s founders chose the story and read every line. **Figure: India gets in-country processing for one OpenAI model, at a 10% premium** *Terms of OpenAI's API regional processing in India, as published on 12 November 2026* 1 model GPT-6 Luna, through the Responses and Chat Completions APIs OpenAI +10% on regional processing: US$0.11 in and US$0.55 out per million tokens OpenAI ZDR Zero Data Retention must be active, on OpenAI's approval OpenAI 6 of 10 OpenAI residency regions that also process requests in-region TechBag count Source: OpenAI data-controls guide and GPT-6 Luna model page; region count is TechBag's · as of 12 November 2026 ## Where the data sits, and where the model runs Data residency and inference residency are different promises. Residency at rest means customer content is stored in the chosen region where an endpoint has to keep data, as for batch jobs. Inference residency means the model also does its work there. OpenAI lists ten regions: Canada, Singapore, South Korea and the UK store only, and there OpenAI “may also process and temporarily store Customer Content outside of the Region”. The US, Europe, the UAE, Australia, Japan and now India also process; Australia, Japan and India share the one-model limits.[1] Two things never move. System data, meaning account details, metadata, usage and billing records, “may be processed and stored outside the selected region”. And by default OpenAI keeps abuse-monitoring logs, which can contain prompts and responses, for up to 30 days; Zero Data Retention keeps customer content out of them. Any region outside the US needs OpenAI’s approval and a Modified Retention amendment.[1] GPT-6 Luna lists at US$0.10 per million input tokens and US$0.50 per million output tokens before the premium.[2] ## Why Indian banks ask for it The Reserve Bank of India’s circular of 6 April 2018 requires that “the entire data relating to payment systems” be stored “in a system only in India”, including end-to-end transaction details.[4] Its FAQs of 26 June 2019 apply this to banks and to the vendors they engage, and allow processing abroad only if the data is deleted there and brought back within one business day or 24 hours, whichever is earlier.[5] The rule covers payment data, not every record a bank holds. A model call made inside India keeps that data out of the foreign-processing case. OpenAI is not the first to offer this. Since 29 September, Amazon Bedrock has served Anthropic’s Claude Opus 5, Claude Sonnet 5 and Claude Haiku 4.5 with an India profile that keeps inference between AWS’s Mumbai and Hyderabad regions.[6] For a model built in India, our note on [Sarvam’s Prism](https://www.thetechbag.com/blog/sarvam-prism-decision-model-4-rupees-million-tokens) gives its price in rupees. Our viewWhat to check before moving a workload Whether GPT-6 Luna can do the job without hosted tools such as web or file search; whether your organisation has Zero Data Retention approved; whether a 10% higher bill fits; and that billing and usage records still leave India. How we did this Figures as of 12 November 2026, 8:30pm IST This piece was researched and drafted with the help of Claude, an AI model made by Anthropic, which is named in it as a rival offering. Terms are from OpenAI’s data-controls guide and GPT-6 Luna page, which carry no date; OpenAI’s API changelog had no India entry, so the launch date is Inc42’s. Our count: ten regions in OpenAI’s table, six with processing. Our arithmetic: US$0.10 and US$0.50 plus 10%. Prices are in US dollars before Indian taxes. The RBI rule is quoted from its circular and FAQs; whether a given workload falls under it is for each bank to read. Spotted an error? Write to us; corrections are published at the foot of the piece. Sources 1. [OpenAI: Data controls in the OpenAI platform (data residency controls; abuse monitoring)](https://developers.openai.com/api/docs/guides/your-data) — undated page, as published 12 November 2026 2. [OpenAI: GPT-6 Luna model page (pricing; regional processing premium)](https://developers.openai.com/api/docs/models/gpt-6-luna) — undated page, as published 12 November 2026 3. [Inc42: OpenAI Brings API Inference Residency To India To Deepen Enterprise AI Adoption](https://inc42.com/buzz/openai-brings-api-inference-residency-to-india-to-deepen-enterprise-ai-adoption/) — 12 November 2026 4. [Reserve Bank of India: Storage of Payment System Data, RBI/2017-18/153, DPSS.CO.OD No.2785/06.08.005/2017-2018](https://www.rbi.org.in/Scripts/NotificationUser.aspx?Id=11244&Mode=0) — 6 April 2018 5. [Reserve Bank of India: FAQs, Storage of Payment System Data](https://www.rbi.org.in/Scripts/FAQView.aspx?Id=130) — 26 June 2019 6. [AWS Machine Learning Blog: Amazon Bedrock expands Claude model availability to in-country inferencing in India](https://aws.amazon.com/blogs/machine-learning/amazon-bedrock-expands-claude-model-availability-to-india-cross-region-inference/) — 29 September 2026 **Disclosure.** TechBag is a software reseller: it earns when companies buy through it, including from vendors named here. The analysis is TechBag’s own; no vendor paid for or saw this piece before publication. Facts are labelled reported, TechBag count or our view. **Read this next:** [Enterprise AI hit a US$177bn run rate, Menlo estimates; Anthropic and OpenAI take 86%](https://www.thetechbag.com/blog/enterprise-ai-spend-177bn-anthropic-openai-86-percent) — The market behind the API: enterprise AI spending at a US$177bn run rate, Menlo estimates, with Anthropic and OpenAI taking 86% of model APIs. Go deeper on TechBag - [Sarvam’s Prism costs ₹4 per million input tokens](https://www.thetechbag.com/blog/sarvam-prism-decision-model-4-rupees-million-tokens) — an Indian model's API price, in rupees - [India’s large data centres run 90% full](https://www.thetechbag.com/blog/india-data-centres-90-percent-full-crisil) — the capacity in-country inference runs on - [SBI’s leak set three reporting clocks](https://www.thetechbag.com/blog/sbi-leak-three-clocks-six-hours-72-hours) — the RBI, CERT-In and DPDP duties an Indian bank already runs