Silvia Outperforms Every Major AI Model on Tax and Open Sources the Benchmark Test

• Silvia outperformed every frontier and open-source model evaluated across tax-related topics, achieved by a team a fraction of the size of the major AI labs it beat
• The breakthrough came from specialized architecture on proprietary data — a blueprint any domain-focused team can follow
• Silvia is open sourcing the results, the evaluation questions, and the framework itself

NEW YORK--(BUSINESS WIRE)--ProCap Financial, Inc. (Nasdaq: BRR), the first publicly traded agentic finance firm, today published research proving that Silvia, its AI agent lab focused exclusively on finance, outperformed every leading frontier and open source model evaluated on tax-related topics and questions.

Silvia’s team evaluated seven AI products across ten expert-level tax scenarios covering federal tax law and the tax codes of California, Texas, North Carolina, New York, and Florida. The responses for each AI product were retrieved from its respective knowledge sources, then independently scored by a blind judge on a 1–10 scale for factual accuracy under four escalating levels of verification.

Silvia, which retrieves answers from a proprietary library of primary tax law, earned the highest overall score (8.73), outperforming the next-best product, Claude Desktop (8.40), and ranking first across every product tested. You can read the full research report here.

Why this matters beyond tax

The approach is replicable. A small team with the right proprietary data and the right architecture can beat a general-purpose frontier product in a chosen domain. The Silvia team hopes this approach becomes the blueprint for other application-specific products.

“There is an unresolved question among institutional investors about whether general-purpose model providers will capture most of the value in AI or whether domain-specific systems will take share,” said Anthony Pompliano, Chairman and CEO of ProCap Financial. “That argument has been theoretical until now. The largest AI labs on earth employ more researchers than we have employees, and Silvia beat all of them on tax with a team of only four AI engineers working on this problem.”

The benchmark is available

ProCap Financial is releasing not only the results of the evaluations, but the evaluation questions and the framework used to produce them. Publishing the test lets anyone reproduce it, contest it, or run their own model against it.

Silvia now has more than $50 billion in assets connected to its platform across more than 20,000 users.

About ProCap Financial

ProCap Financial is the first publicly traded agentic finance firm. The Company’s mission is to help independent investors make money. Products include Silvia, an AI research lab exclusively focused on finance, and ProCap Insights, the first agentic research offering in finance. Founded in 2025, the Company raised more than $750 million from leading investors and is traded on Nasdaq under the symbol BRR. Visit www.procapfinancial.com for more information.

No Tax Advice

The research described in this release is for informational purposes only and does not constitute tax, legal, investment, or accounting advice. Silvia does not provide tax advice. Benchmark results reflect internal testing under controlled conditions and are not a guarantee of accuracy or performance in any individual case. Tax outcomes depend on individual facts and circumstances and on laws that are subject to change. Users should consult a qualified tax professional regarding their individual circumstances.

Forward-Looking Statements

This press release contains forward-looking statements within the meaning of the Private Securities Litigation Reform Act of 1995. These statements are based on current expectations and are subject to various risks and uncertainties that could cause actual results to differ materially. Such risks include those described under “Risk Factors” in most recent SEC filings. We caution readers not to place undue reliance on forward-looking statements, which reflect our views only as of the date of this release. We undertake no obligation to update such statements, except as required by law.