On November 5 Julius Research CEO rahul (@0interestrates; Julius AI) released DeepFinance: 21 ultra long-horizon PE/IB tasks built from anonymized real deal data rooms (~20–30 human hours each, mean ~27.1h; ~84 rubric criteria). Official leaderboard: Opus 5.5 (max) ~54.8%, Sonnet 5.5 ~54.6%, GPT-6 Astra ~54.2%. Focus is auditable trajectories, not final-number checks alone.
Key Takeaways
- ✓Primary source: Julius Research CEO @0interestrates post + official leaderboard/blog (first-hand lab lead release, not a separate corporate PR account)
- ✓Scale: 21 tasks; mean ~27.1 human hours; ~84 rubrics/task; ~19 files per data room
- ✓Top (lab-reported): Opus 5.5 max ~54.8%; Sonnet 5.5 ~54.6%; GPT-6 Astra xhigh ~54.2%
- ✓Cost (official table): Opus ~$8.63/task; Astra ~$4.55; GPT-6.1 Sol ~$0.81 at ~52.9%
- ✓Gap filled: moves beyond doc QA / final-number checks toward auditable trajectories under conflicting evidence
Key Decision Metrics at a Glance
Turn your technical choice into a development budget
Compare 40 dev plans & simulate token costs vs $20/mo subscriptions
Project Links & Resources
Direct AccessIn-Depth Technical Analysis
Core Background & Industry Pain Points
Most finance evals stop at document QA or checking a final number, missing the assumptions, conflicting evidence and auditable trajectories in PE/IB diligence. On November 5 Julius Research CEO rahul (@0interestrates; Julius is their AI data-scientist product) first-hand released DeepFinance, an ultra long-horizon benchmark for private-equity and investment-banking workflows.
Architecture Highlights & Internals
Official site/blog: 21 tasks built from anonymized real deal data rooms (~19 files on average), authored/reviewed by experts with backgrounds at Evercore, J.P. Morgan, UBS and institutions such as HBS, Wharton and Columbia. Mean human time ~27.1 hours per task (about 20–30 hours); ~84 rubric criteria each. Scoring grades end-to-end workflows against rubrics, not only final numbers. Harness: 10 attempts per task with mean scores and 95% bootstrap CIs.
Authoritative Benchmarks & Measured Scores
Julius leaderboard (updated Nov 5, 2026; lab-reported): Claude Opus 5.5 (max) ~54.8%, Sonnet 5.5 (max) ~54.6%, GPT-6 Astra (xhigh) ~54.2%, GPT-6.1 Sol (xhigh) ~52.9%; costs about $8.63 / $4.55 / $0.81 per task for Opus / Astra / Sol. Most models remain far from reliable diligence; many open models sit near 20–35%. Not AICoder re-runs.
Developer Hands-on Guide
See juliusresearch.ai/deep-finance and the blog “DeepFinance: Can AI diligence a $100MM deal?”. Useful for auditing long-horizon finance agents on assumption quality and evidence alignment; tasks/data rooms are not claimed as public downloads. Cite Julius Research figures and separate “numerically correct” from “auditable trajectory”.
Benchmark side-by-side against alternatives, or calculate monthly token cost vs subscription break-even.
Discussion & Comments
0Sign in to join the discussion
Connect with AI developers to exchange benchmark insights.