Introduction Candida haemulonii is now considered to be a complex of two species and one variety : C. haemulonii , Candida duobushaemulonii and the variety Candida haemulonii var. vulnera . The correct identification (ID) of these species is clinically relevant, since resistance to azole derivatives has been reported and amphotericin B has poor in vitro activity against C. duobushaemulonii isolates. Initial evaluation using MALDI-TOF MS for the ID of the C. haemulonii complex species provided promising results, but the discrimination of C. haemulonii from the variety vulnera was problematic, and the performance of the VITEK MS remains unevaluated. Objectives Evaluation of the MALDI-TOF MS performance for the ID of C. haemulonii species ID with two platforms and its’ databases and softwares. Methods 14 C. haemulonii sensu stricto , 9 C. haemulonii var. vulnera and 10 C. duobushaemulonii from hospitals belonging to the Medical school from the University of Sao Paulo were analyzed (ID by ITS1 sequence analysis). A set of reference strains from the CBS-KNAW collection was also included : C. haemulonii CBS5149 T, C. duobushaemulonii CBS7798 T, C. duobushaemulonii CBS7799, and for specificity control, the close related species Candida pseudohaemulonii (CBS10004 and CBS12370). The isolates/strains were cultured on Sabouraud plates and incubated for 48 h at 30 °C before MALDI-TOF MS analysis. Protein extraction protocol with ethanol and formic acid 70 % was carried out according to the Bruker's recommendations. Measurements were performed on a Microflex LT and its standard database (database 3.3.1), and VITEK MS instrument equipped with both IVD and RUO (SARAMIS) databases. To differentiate C. haemulonii (target group) from the variety vulnera (control group), mass spectra analysis models were created with the ClinProTools 3.0 software (Bruker). For each model, the recognition capability (RC) and cross validation (CV) percentage was generated. For single peak analysis, the AUC of each peak for the discrimination of target group from control group was directly obtained from the ClinProTools software. For the peaks with the highest AUC, the detection performances were checked using FlexAnalysis 3.4 (Bruker Daltonics). The signal-to-noise ratios (SN) of the peaks were exported to SPSS 18.0, ROC curves were constructed, and their optimal cut-off values were determined. Results All C. haemulonii sensu stricto isolates/strain had correct species assignment by both Bruker and VITEK MS IVD databases. For the species C. duobushaemulonii, the Bruker database gave correct species ID for 75 % of the isolates/strains. VITEK MS IVD analysis misidentified all C. duobushaemulonii and C. pseudohaemulonii isolates/strains as C. haemulonii (99 % of confidence level). The VITEK MS SARAMIS was unable to give genus/species ID for all isolates/strains. Discrimination of the isolates belonging to the C. haemulonii sensu stricto and to the variety vulnera was not possible with all systems and databases. The ClinProTools models showed values of CV and RC above ≥90 % for the discrimination of target group from control group. The most discriminative peaks were 5107, 6878 and 13,750 m/z (AUC >0,9), with sensibility and specificity of 88,6 %, 80,8 %, 85,8 %, and 96,6 %, 92,9 %, 93,8 %, respectively. The SN cut-off values from the peaks 5107, 6878 and 13,750 m/z for the discrimination of the target group from the control group were 2,5, 3,9 and 6,3, respectively.