Free fatty acids (FFAs) in fat are important indicator of raw milk quality. Result reliability of FFAs is important. Aim was to verify MIR–FT (mid infrared spectroscopy with Fourier’s transformations) method for its calibration to determine FFAs, time stability of MIR–FT FFA calibration and calibration levelling in laboratory networks. Reference (RE) milk samples (1 set = 8) were prepared according to CSN 57 0533 (FFAs in mmol.100g−1 of fat). MIR–FT instruments were: 1 LactoScope FTIR (DE); 2 Bentley FTS (BE); 2 MilkoScan FT 6000 (FO). 3 calibrations of MIR–FT (5) in 3 laboratories were performed. Bulk milk samples came from 4 herds and 2 breeds. These 4 samples were used for calibration in native and modified form. Modification increased FFAs by cca 100%. Calibration set had 8 samples. 1 between calibration interval was checked monthly by proficiency testing (PT). PT set had 10 samples. 5 samples were with native milk and 5 had modified fat content, lower and higher. Maximal value of difference variability for calibration quality validation is x (sd of difference MIR–FT and RE) plus 1.64 multiple of sd (on 95% level), 1.0613 mmol.100g−1. Mean validation correlation coefficient (r) between MIR–FT and RE results was 0.802 ± 0.082 (P < 0.001), from 0.666 to 0.945. Minimal value at calibration is x minus 1.64 multiple of sd (0.668). Correlations between MIR–FT results were higher by 8.4% (0.869 (P < 0.001) > 0.802). Example PT with 10 and 5 milk samples had similar results of r 0.887 and 0.953 (P < 0.001 and P < 0.05). There is possibility to construct a levelling programme for calibrated instruments. Some equation between PT reference and instrumental values could correct MIR–FT results for their better comparability.