BACKGROUND: Free thyroxine (FT4) and free triiodothyronine (FT3) measurements are useful in the diagnosis and treatment of a variety of thyroid disorders. The IFCC Scientific Division established a Working Group to resolve issues of method performance to meet clinical requirements. METHODS: We compared results for measurement of a panel of single donor sera using clinical laboratory procedures based on equilibrium dialysis-isotope dilution-mass spectrometry (ED-ID-MS) (2 for FT4, 1 for FT3) and immunoassays from 9 manufacturers (15 for FT4, 13 for FT3) to a candidate international conventional reference measurement procedure (cRMP) also based on ED-ID-MS. RESULTS: For FT4 (FT3), the mean bias of 2 (4) assays was within 10% of the cRMP, whereas for 15 (9) assays, negative biases up to -42% (-30%) were seen; 1 FT3 assay was positively biased by +22%. Recalibration to the cRMP eliminated assay-specific biases; however, sample-related effects remained, as judged from difference plots with biologic total error limits. Correlation coefficients to the cRMPs ranged for FT4 (FT3) from 0.92 to 0.78 (0.88 to 0.30). Within-run and total imprecision ranged for FT4 (FT3) from 1.0% to 11.1% (1.8% to 9.4%) and 1.5% to 14.1% (2.4% to 10.0%), respectively. Approximately half of the manufacturers matched the internal QC targets within approximately 5%; however, within-run instability was observed. CONCLUSIONS: The study showed that most assays had bias largely correctable by establishing calibration traceability to a cRMP and that the majority performed well. Some assays, however, would benefit from improved precision, within-run stability, and between-run consistency.