OpenAI has released details on Terminal X, an AI system that achieves an 88.6% match rate with senior investor judgment when powered by GPT-6.1. The system appears designed for financial oversight, specifically evaluating whether investment outputs stay aligned with a fund's stated mandate.
The core of Terminal X is its Eval Agent, which grades outputs against a 54-criterion rubric. This structured approach allows the system to detect subtle compliance issues that might otherwise be missed. In testing, the Eval Agent caught 25% more deviances from a fund's mandate compared to baseline methods, according to OpenAI.
While the source highlights strong performance metrics, it does not provide direct comparison with other models or human-only processes. The improvement in mandate compliance suggests that AI-based evaluation could become a useful check on investment decisions, though broader validation would be needed to confirm its reliability across diverse funds and market conditions.