Gemcitabine (GEM) is first-line therapy for pancreatic cancer but has limited efficacy in most cases. Nanoparticle-albumin bound (nab)-paclitaxel is becoming first-line therapy for pancreatic cancer, but also has limited efficacy for pancreatic cancer. Our goal was to improve the treatment outcome in patient-like models of pancreatic cancer. We previously established patient-derived orthotopic xenografts (PDOX) pancreatic cancers from two patients. The pancreatic tumor was implanted orthotopically in the pancreatic tail of nude mice to establish the PDOX models. Five weeks after implantation, 50 PDOX mouse models were randomized into five groups of 10 mice for each pancreatic cancer PDOX: untreated control; GEM (100 mg/kg, i.p., once a week for 2 weeks); GEM + nab-PTX (GEM: 100 mg/kg, i.p., once a week for 2 weeks, nab-PTX: 10 mg/kg, i.v., twice a week for 2 weeks); S. typhimurium A1-R (5 × 10 CFU/100 μl, i.v., once a week for 2 weeks); GEM + S. typhimurium A1-R (GEM: 100 mg/kg, i.p., once a week for 2 weeks, S. typhimurium A1-R; 5 × 10 CFU/100 μl, i.v., once a week for 2 weeks). GEM + nab-PTX was significantly more effective than GEM alone in one PDOX model (p = 0.0004), but there was no significant difference in the other PDOX model. The combination of GEM + S. typhimurium A1-R regressed both PDOX models. These results show S. typhimurium A1-R can overcome the ineffectiveness or partial effectiveness of GEM in patient-like models of pancreatic cancer and demonstrate clinical potential for this combination.