Finance Agents Benchmark: leaderboard
SecondState's FAB: an agent works through a synthetic company data room (160 documents) to answer 50 financial due-diligence tasks. A task passes only when every rubric criterion passes, judged by an LLM; scored by task pass rate over three trials, higher is better (2026).
Metric: Task pass rate (%) over 50 due-diligence tasks x 3 trials; a task passes only when every rubric criterion passes (GPT-6 Luna judge). Source: github.com. Saturation forecast: Around April 2028. 4 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | GPT-6 Sol (Medium) | 58.7 | #25 (GPT-6 Sol) |
| 2 | GPT-6 Luna (Medium) | 50.7 | #95 (GPT-6 Luna) |
| 3 | GLM-5.3 Flash (Medium) | 47.3 | #122 (GLM-5.3 Flash) |
No result here: #1 Claude Fable 5.2, #2 Gemini 4 Carbon, #5 Claude Opus 5.5.
Interactive version: theaggregate.ai/benchmark?slug=finance-agents-benchmark · How It Works · Data refreshed daily, snapshot 2026-11-13.