Background: Acute febrile illness is a common presentation for patients at hospitals globally. Assays that can diagnose a variety of common pathogens in blood could help to establish a diagnosis for targeted disease management. We aimed to evaluate the performance of the BioFire Global Fever Panel (GF Panel), a multiplex nucleic acid amplification test performed on whole blood specimens run on the BioFire FilmArray System, in the diagnosis of several pathogens that cause acute febrile illness. Methods: We did a prospective, multicentre, cross-sectional diagnostic accuracy study to evaluate the GF Panel. Consenting adults and children older than 6 months presenting with fever in the previous 2 days were enrolled consecutively in sub-Saharan Africa (Ghana, Kenya, Tanzania, Uganda), southeast Asia (Cambodia, Thailand), central and South America (Honduras, Peru), and the USA (Washington, DC; St Louis, MO). We assessed the performance of six analytes (chikungunya virus, dengue virus [serotypes 1–4], Leptospira spp, Plasmodium spp, Plasmodium falciparum, and Plasmodium vivax or Plasmodium ovale) on the GF Panel. The performance of the GF Panel was assessed using comparator PCR assays with different primers followed by bidirectional sequencing on nucleic acid extracts from the same specimen. We calculated the positive percent agreement and negative percent agreement of the GF Panel with respect to the comparator assays. This study is registered with ClinicalTrials.gov, NCT02968355. Findings: From March 26, 2018, to Sept 30, 2019, 1965 participants were enrolled at ten sites worldwide. Of the 1875 participants with analysable results, 980 (52·3%) were female and the median age was 22 years (range 0–100). At least one analyte was detected in 657 (35·0%) of 1875 specimens. The GF Panel had a positive percent agreement for the six analytes evaluated as follows: chikungunya virus 100% (95% CI 86·3–100), dengue virus 94·0% (90·6–96·5), Leptospira spp 93·8% (69·8–99·8), Plasmodium spp 98·3% (96·3–99·4), P falciparum 92·7% (88·8–95·6), and P vivax or P ovale 92·7% (86·7–96·6). The GF Panel had a negative percent agreement equal to or greater than 99·2% (98·6–99·6) for all analytes. Interpretation: This 1 h sample-to-answer, molecular device can detect common causative agents of acute febrile illness with excellent positive percent agreement and negative percent agreement directly in whole blood. The targets of the assay are prevalent in tropical and subtropical regions globally, and the assay could help to provide both public health surveillance and individual diagnoses. Funding: BioFire Defense, Joint Project Manager for Medical Countermeasure Systems and US Army Medical Materiel Development Activity, and National Institute of Allergy and Infectious Diseases.