Infrastructure

OpenAI Lets Japanese Customers Run GPT-6 Luna API Processing and Inference Inside Japan

· 6 min read · Infrastructure

OpenAI Lets Japanese Customers Run GPT-6 Luna API Processing and Inference Inside Japan

OpenAI has made generally available an option that lets companies and developers in Japan have API requests processed and model inference run inside the country, starting with its GPT-6 Luna model, Impress Watch reported on November 11. The feature covers text and image inputs with text outputs and is open to customers that meet the usage conditions. In May 2025, OpenAI had already announced data residency for Japan and other parts of Asia, allowing eligible customer data to be stored in the country. The new step extends that from storage to the processing of requests and the generation of answers.

Access is tightly scoped. Only the Responses API and the Chat Completions API are covered, and customers must first receive OpenAI’s approval for zero data retention (ZDR) and switch it on for those APIs. They then create a new API project with Japan selected as both the storage and processing location, and call a Japan-specific endpoint. Only the Standard pay-as-you-go service tier is supported. The ChatGPT and Codex apps, models other than GPT-6 Luna, voice, image generation and OpenAI-hosted tools are excluded.

What it means in practice

OpenAI Lets Japanese Customers Run GPT-6 Luna API Processing and Inference Inside Japan — contextual photo

The change is aimed at sectors where data must be handled in Japan, such as finance, telecoms, healthcare and public administration, for tasks like summarising documents, drafting answers and supporting analysis. Impress Watch also noted the limits. In-country storage and processing applies only to defined customer content, not to system data such as account details, usage records or billing information, and tools or external services run by the customer fall outside the scope, so OpenAI does not guarantee that every step stays in Japan. OpenAI said it is offering the same capability in Australia and India.

Keeping AI workloads on Japanese soil has become a recurring theme, from NTT and SoftBank’s in-Japan AI security services to NVIDIA’s new Japanese synthetic datasets for finance and privacy work. What to watch is whether OpenAI extends domestic processing to more models and to its ChatGPT products, how quickly banks and public bodies adopt the Japan endpoint, and how the offer compares with the domestic options promoted by Japanese cloud providers.

← Back to AI4Japan