Agents & tool use

DeepFinance

by Julius Research (Drew Hahn, Tyler Shibata, Jackson Grove, Rahul Sonwalkar et al.)

What is DeepFinance? DeepFinance is an agents & tool use eval made by Julius Research (Drew Hahn, Tyler Shibata, Jackson Grove, Rahul Sonwalkar et al.). An ultra long-horizon PE and IB workflow benchmark of 21 anonymised deal tasks (about 27.1 human hours and about 84 rubric criteria each) with data rooms of spreadsheets, emails, PDFs and notes. Graders score the deliverable and the agent's trajectory on output, workflow and judgment criteria written by PE/IB practitioners. Models run in their native harnesses; Julius's 5 Nov 2026 lab run reports a top mean rubric score of 54.8% (Claude Opus 5.5 max), with the top three within about one point. Tasks stay private; outside teams cannot rerun the official set.

Page last updated 2026-10-18 · Evals Review

Editor note

An ultra long-horizon PE and IB workflow benchmark of 21 anonymised deal tasks (about 27.1 human hours and about 84 rubric criteria each) with data rooms of spreadsheets, emails, PDFs and notes. Graders score the deliverable and the agent's trajectory on output, workflow and judgment criteria written by PE/IB practitioners. Models run in their native harnesses; Julius's 5 Nov 2026 lab run reports a top mean rubric score of 54.8% (Claude Opus 5.5 max), with the top three within about one point. Tasks stay private; outside teams cannot rerun the official set.

What it tests
Whether agents can finish multi-hour private-equity and investment-banking diligence with auditable assumptions and sourced numbers.
Original source
www.juliusresearch.ai/blog/deep-finance
Last updated
Released 2026-11-05
License
Not checked yet — see source

Facts come only from the linked source (checked 15–16 Oct 2026). Credit belongs to the creators. We don't score or rank evals: how we review.

Reviews

No reviews yet.

Reviews are written by people who sign in with GitHub, one per account per eval. We can't confirm that each reviewer used the eval; we remove reviews that break our rules, never just for being negative, and we never add, buy or edit reviews. Averages, when shown, come only from these reviews.

Sign in with GitHub to review

Related agents & tool use evals

  • GAIA — General assistant skills: browsing, file handling, multi-step reasoning.
  • WebArena — Autonomous web navigation and task completion.
  • OSWorld — Computer-use agents operating real operating systems and apps.
  • Terminal-Bench — Agents working in the shell end to end.

All Agents & tool use evals · Comparison table

Discussion

No one has discussed DeepFinance yet. Have you run it? Share what it caught, what it missed or how you set it up.

Sign in with GitHub to comment

Be specific and kind. No promotion or affiliate links. Community rules · All discussions