OpenAI announced the establishment of OpenAI Japan on 14 April 2024, marking its first Asian office and third international location after London and Dublin [source]. The Tokyo-based entity will serve enterprise customers across Japan and introduce a GPT-4 custom model optimised for the Japanese language.

The custom model, developed in collaboration with an undisclosed Japanese research institution, is described as offering improved performance on Japanese text compared to existing GPT-4 variants. OpenAI stated the model achieves faster processing speeds and better contextual understanding of Japanese grammar and cultural nuances, though no independent benchmarks were provided at announcement.

Tadao Nagasaki, formerly of Amazon Web Services Japan, was named president of OpenAI Japan. The company disclosed partnerships with Daikin Industries, Rakuten Group, and the Tokyo Metropolitan Government as early enterprise adopters. These organisations will deploy GPT-4 and the Japan-optimised variant for customer service automation, content generation, and administrative workflows.

The announcement follows OpenAI's pattern of regional expansion through localised model variants. A Korea-focused office was reportedly under consideration but not confirmed. OpenAI did not disclose the training data composition for the Japanese model or whether it involved additional fine-tuning beyond GPT-4's base training.

The Tokyo office will house sales, support, and partnership teams. OpenAI indicated plans to hire local engineering staff but did not specify whether model development would occur in Japan. The company stated it will comply with Japan's Act on the Protection of Personal Information and work with regulators on AI governance frameworks.

No technical documentation or API access details for the Japanese-optimised model were released at announcement.