Japan Research Institute licenses PLaMo, deploys PLaMo 3.0 Prime from October 2026

The agreement follows about six months of evaluation by Japan Research Institute, which tested PLaMo against real business scenarios before deciding to adopt it. The firm wanted a private LLM that could be operated stably and at a high level inside its own environment without sending confidential information outside, and it also valued PLaMo's Japanese-language performance, transparency of training data, and the ease of working with the developer on model improvements.

Under the agreement, the initial rollout will focus on two areas: system development and operations, including review support for design documents and source code, investigation of impact when failures occur, and handling inquiries about specifications and past response histories for long-running systems; and internal back-office work such as searching and responding to internal regulations, drafting minutes and reports, and translation. In parallel, Japan Research Institute will verify additional training of PLaMo using its own data, aiming to build a model that reflects its business know-how and can be used widely across operations.

Preferred Networks positions this as a leading example of domestic AI taking root in financial operations, and plans to continue providing successor models while working closely with Japan Research Institute to support the expansion of its AI use. PLaMo is a fully scratch-developed domestic generative AI foundation model, and PLaMo 3.0 Prime is based on a pre-training model obtained through joint research with NICT.