Terminal X says GPT-6.1 Sol matched investor reviews in 88.6% of test cases

Terminal X said its Eval Agent agreed with a senior investment reviewer on 88.6% of 500 evaluated judgments using GPT-6.1 Sol. An October 20 OpenAI case study put the next-best model at 86.4%, without identifying that model.

The agent evaluates investment work using 54 criteria covering evidence, reasoning and consistency with a firm's strategy. Terminal X said GPT-6.1 Sol detected about 25% more departures from fund mandates and reduced costs per query by roughly 30% against the comparison model. These are company-reported evaluation results, not measurements of investment returns.

Terminal X's Capital Memory system records how investment teams analyze opportunities and turns those processes into customized agents. Failed checks generate corrections that inform later workflows or tests. The company plans to extend these records from investment desks to individual professionals, capturing how each person applies judgment.