Preferred Networks launches PLaMo 3.1 Prime AI model

Preferred Networks (PFN) has released PLaMo 3.1 Prime, a new flagship AI model developed in Japan. The model features a “Deep Research” function capable of autonomously creating research plans and generating reports with citations within minutes or hours. This capability is designed to support corporate research tasks, such as market analysis and technical trend monitoring.

The model supports on-premises deployment, allowing companies to integrate it with internal document search systems. This enables research that combines internal data with web-based information or focuses exclusively on private documents in secure environments. PFN reports that the model shows improved factuality and faithfulness compared to its predecessor, with a 15% average score improvement in hallucination-related benchmarks.

PFN has introduced a free tier for the PLaMo API with usage limits. For the PLaMo Chat service, a promotional campaign allows users to try the Deep Research feature for free up to twice a day until November 30. While the model outperformed competitors like Alibaba’s Qwen and Google’s Gemma in safety benchmarks, it trailed in areas such as STEM, coding, and web-search-based performance.