Neon, a provider of serverless Postgres databases, collaborated with Castform, a company offering a platform for further training open-source AI models for enterprise use, to post-train an open-source model with 4 billion parameters using reinforcement learning. They reported achieving "accuracy equivalent to OpenAI's GPT-5.6 Sol in search tasks at 1/100th the cost." Traditionally, "embedded search" has been used for searching internal company documents, but with the proliferation of AI agents, there is a growing trend toward "agent-based search," where the AI model independently determines the search query, reviews results, and performs the next search. This approach handles more complex questions but increases processing time and costs because a high-performance model is called for each search. According to Neon, a typical search request using GPT-5.6 Sol takes more than 10 seconds and costs approximately $0.03 (about 4.5 yen) per request. Castform trained a relatively small open-source model to learn how to find necessary information from specific data, with Neon providing document storage and search functionality. The model is evaluated on whether it found correct documents, cited appropriate passages, and gave correct final answers. Many companies lack question-and-answer data for AI training, but have accumulated internal documents, product information, support articles, customer interactions, and business databases. Castform automatically generates question-and-answer combinations based on this existing data, such as creating questions from travel regulations and training the model with