Terminal X reports 88.6% agreement with senior investors using GPT-6.1
Terminal X says its agents, using OpenAI’s GPT-6.1 models, matched senior investor judgment 88.6% of the time across 500 judgments.

Key takeaways · 3
- 01
Terminal X reports 88.6% agreement with senior reviewers across 500 judgments.
- 02
The same workflows ran at approximately 30% lower cost per query than with the next-best model.
- 03
Terminal X says customers own their agents and can deploy them on-premise or in their own cloud.
What the benchmark reports
Terminal X says its system matched senior investor judgment 88.6% of the time across 500 judgments.[1][2] Blockchain.News reports that the next-best model achieved 86.4% agreement, while OpenAI says GPT-6.1 Sol delivered 220 basis points higher agent agreement with human reviewers than that model.[1][2] The reported comparison is useful as a benchmark, but it does not establish that an agent can replace a firm’s investment review. Blockchain.News cautions that the results were vendor-reported and were not presented as an independent audit.[2]
How the agents are configured
Terminal X says its forward-deployed engineers work with customers’ investment teams to design agent architectures and encode how each firm invests.[1] The company says it builds hundreds of agents customized by asset class and strategy, with sub-agents for each investment stage.[1] Its Eval Agent grades outputs against a 54-criterion rubric covering source selection, analytical coherence and alignment with a firm’s house view.[1] Terminal X also says production queries are auditable through a trail recording steps, retrieved sources, the model used and reasoning.[1]
Model roles and reported efficiency
OpenAI says GPT-6.1 Sol and GPT-6 Luna handle reasoning inside Terminal X’s agents.[1] Terminal X says Luna is used for high-volume steps where token efficiency matters more than depth, while OpenAI says Sol’s long-context reasoning and token efficiency support deeper evaluation pipelines.[1] OpenAI reports that Sol helped the Eval Agent catch about 25% more recommendations that strayed from a fund’s mandate.[1] The same workflows ran at approximately 30% lower cost per query than with the next-best model.[1] These figures describe the reported workflows, not a guarantee of equivalent results in other settings.
Deployment and reported savings
Terminal X says customers own their agents and may deploy them on-premise or in their own cloud.[1] The company says its agents process investment documents in English, Korean, Japanese and Chinese.[1] OpenAI says one private-equity customer reported $2.4 million in annualized savings from automated due-diligence and investment-memo workflows.[1] That figure is a customer-reported result for one customer, rather than a stated outcome across Terminal X’s customer base. Terminal X’s announcement says it announced its collaboration with OpenAI on September 1, 2026.[3]
Investment teams considering AI agents can compare the reported accuracy, cost and mandate-alignment figures with their own review requirements. They should also weigh deployment choices, auditability and the vendor-reported nature of the results before treating the benchmark as evidence of performance in their own workflows.
Why it matters
Bring this to your team — one session a day, built for your industry.
kju.ai is built for teams: daily sessions, team analytics, and rollout support for your organization.
For teamsHow this developed
21 October 2026
Terminal X reports 88.6% agreement with senior investors using GPT-6.1