




Backed by Founders Fund, Symbal has spent the last two years building specific intelligence for the enterprise — including nine of the leading private capital firms with $100B in deployed capital. Time and time again, these financial firms continue to partner with us because frontier models still fail on associate-grade work.
While there are multiple benchmarks that probe agent capabilities in finance (vals.ai's Finance Agent V2, APEX Agents, PR Bench - Finance), we do not find them representative of the types of tasks that are important to our customers. These benchmarks often test how models reason over (1) public filings or (2) on real private-models but with synthetically generated inputs.
However, real analyst work happens in private Data Rooms. Each data room spans hundreds of files, encoding the workstreams of not only the investment committee but also the accounting, consulting, and legal partners involved in every deal that a private capital firm diligences.
To reflect this reality, we built DealBench: the first-of-its-kind finance eval suite. Every task is one a real associate performed on a real transaction — and we hold the complete record: every document and version, spanning the banking, accounting, and legal workstreams from pitch to close.
Evaluates key-person retention risk in a growth-equity target by analyzing employment agreements, vesting schedules, and change-of-control provisions across the data room.
Experts from Insight Partners, Partners Group, WestCap
GPT 5.5
Opus 4.8
Opus 4.7Maps required third-party consents, lender waivers, and tenant notifications across portfolio assets to flag blockers and accelerate a real-estate close-out process.
Experts from Shore Capital, Houlihan Lokey, Audax Private Equity
GPT 5.5
Opus 4.7
Opus 4.8