Identification and antimicrobial susceptibility testing (AST) are critical steps in the management of bloodstream infections. Our objective was to evaluate the performance of the Accelerate Pheno™ System, CE v1.2 software, for identification and AST of Gram-negative pathogens from positive blood culture bottles. A total of 104 bottles positive for Gram-negative bacteria collected from inpatients throughout our institution were randomly selected after Gram staining. The time-to-identification and AST results, and the raw AST results obtained by the Accelerate Pheno™ system and routine techniques (MALDI-TOF MS and VITEK®2, EUCAST guidelines) were compared. Any discrepant AST result was tested by microdilution. The Pheno™ significantly improved turn-around times for identification (5.3 versus 23.7 h; p < 0.0001) and AST (10.7 versus 35.1 h; p < 0.0001). Complete agreement between the Accelerate Pheno™ system and the MALDI-TOF MS for identification was observed for 96.2% of samples; it was 99% (98/99) for monomicrobial samples versus 40% (3/5) for polymicrobial ones. The overall categorical agreement for AST was 93.7%; it was notably decreased for beta-lactams (cefepime 84.4%, piperacillin-tazobactam 86.5%, ceftazidime 87.6%) or Pseudomonas aeruginosa (71.9%; with cefepime 33.3%, piperacillin-tazobactam 77.8%, ceftazidime 0%). Analysis of discrepant results found impaired performance of the Accelerate Pheno™ system for beta-lactams (except cefepime) in Enterobacteriales (six very major errors) and poor performance in P. aeruginosa. The Accelerate Pheno™ system significantly improved the turn-around times for bloodstream infection diagnosis. Nonetheless, improvements in the analysis of polymicrobial samples and in AST algorithms, notably beta-lactam testing in both P. aeruginosa and Enterobacteriales, are required for implementation in routine workflow.