The uniquely American phenomenon of mass incarceration plagues the pretrial space. People awaiting trial make up roughly 20% of those held in criminal custody in the United States. Largely overlooked by bail-reform advocates, pretrial detention in the federal criminal system presents a puzzle. The federal system detains defendants at a much higher rate than the states—more than 60% of U.S. citizen-defendants were detained pending trial by federal courts last year. But federal defendants virtually never fail to appear in court, and they are rarely arrested for new crimes while on pretrial release. And unlike state court systems, cash bail is disfavored in federal courts. Most federal defendants who are released pending trial are released on personal recognizance or unsecured bond. This Article argues that the federal experience with pretrial detention— beginning with its historical roots in old English law and ending with the enormous and disparate detention rates that I document today—provides important lessons for those seeking to reform bail in both federal and state law. This Article tackles a critical empirical question: does the modern, broadly discretionary, federal detention regime generate race- or gender- based disparities in pretrial detention? To answer this question, this Article leverages an expansive new dataset that covers more than 300,000 federal defendants sentenced between fiscal years 2002 and 2016. The results are sobering. White defendants are more likely to be released pending trial than otherwise similar Black and Hispanic defendants, and female defendants are more likely to be released pending trial than otherwise similar male defendants. More importantly, a disparity analysis that recognizes the intersectional relationship between race and gender paints a more complex picture. The most extreme racial disparities are among male defendants—where Black men are four percentage points and Hispanic men are six percentage points more likely to be detained than similarly situated white men. Among female defendants, however, racial disparity presents the opposite relationship: Black women are more likely to be released pending trial than Hispanic and white women. Notably, this disparity pattern for women does not appear in other studies of pretrial detention in state courts, where white women are often the most likely race–gender group to be released. Based on this empirical evidence, the Article assesses several possible legal changes to address disparity, including amending the federal Bail Reform Act to allow judges to consider costs in detention decisions, limiting or prohibiting the consideration of dangerousness, expanding appellate review, and narrowing or eliminating statutory presumptions of detention. The Article ends by offering cautionary lessons for states embarking on bail reform.