Sep 2, 2026
Lunon Deep Research ranks first on Deep Research Bench
Lunon Deep Research, the research agent behind every Lunon engagement, has placed first on Deep Research Bench.
Deep Research Bench is an independent evaluation run by FutureSearch. It scores research agents on 91 real-world tasks, the kind an analyst is handed on a Monday morning: find the number, trace the chain of evidence, reconcile sources that disagree, and say what is known and what is not. Each task runs against a frozen snapshot of tens of thousands of web pages, so results are stable and answers are graded against worked-out solutions rather than by taste.
Why this benchmark
Most benchmarks reward fluent summaries. Deep Research Bench rewards being right. Its tasks have known answers, so an agent that writes confidently but cites the wrong figure scores the same as one that gives up.
That is the standard Lunon holds its own work to. A diligence memo is only useful if the numbers survive an investment committee.
What we built
Lunon Deep Research is the research layer inside Lunon’s engagements. It breaks a question into the sub-questions that decide it, gathers evidence across sources, checks claims against each other, and keeps the trail from every number back to where it came from. The same system runs commercial diligence, market assessments, and strategy work for Lunon’s clients.
A public benchmark matters because it is measured by someone else. It does not change how we work: every finding still goes through a Lunon expert before it reaches a client.
What comes next
We will keep submitting as the benchmark evolves, and we will publish what we learn about where research agents still fall short. If you want to see the system on your own question, get in touch.
Start with the decision in front of you.
Tell us what your team needs to understand, evaluate, or deliver.