Objective To compare the outcomes and safety of 23G and 20G vitrectomy for treatment of infectious endophthalmitis. Methods This was a retrospective case study. Sixtyseven eyes of 67 eyes suffering from infectious endophthalmitis with a history of trauma or intraocular operation history were enrolled in this study. They were diagnosed by the examinations of best corrected visual acuity, intraocular pressures, slit lamp microscope, indirect ophthalmoscopy, B-scan ultrasound and CT. There were 49 males (49 eyes) and 18 females (18 eyes). The patients aged from 18 to 72 years with a mean age of (43plusmn;13) years. There were 60 patients (60 eyes) with a history of trauma, 7 patients (7 eyes) with intraocular operation history. The patients were enrolled into 20G vitrectomy group (35 patients, 35 eyes) before December, 2009 and 23G vitrectomy group (32 patients, 32 eyes) after January, 2010 when 23G vitrectomy system was imported in this hospital. Vitreous purulence was taken in all patients at the beginning of the surgery for bacteria and fungal culture and drug sensitivity test. A standard vitrectomy with artificial posterior vitreous detachment followed by internal limiting membrane peeling, and (or) intraocular laser photocoagulation, cryocoagulation, fluidair exchange with intraocular silicone oil or gas tamponade were performed in all cases. Broadspectrum antibiotics and glucocorticoids were used systematically for one week after surgery, but glucocorticoids were not used for fungal infections. The followup was ranged from two to nine months with a mean of (7plusmn;1) months. The surgical time, inflammation situation, visual acuity, intraocular pressure, retinal reattachment rate, iatrogenic retinal hole rate, bulbar conjunctiva scar formation rate, reoperation rate and eye retention situation before and after surgery were comparatively analyzed. Results The mean surgical times were (126plusmn;12) and (89plusmn;12) minutes in 20G and 23G group, which was significantly different (t=3.125, P＜0.05). The major surgery complications were ora serrata dialysis and other iatrogenic retinal breaks, and were occurred in 34 eyes, including 30 eyes (85.71%) in 20G group and 4 eyes (12.50%) in 23G group (chi;2=35.85,P＜0.05). These 4 eyes in 23G group received foreign body removal surgery previously. The inflammation was controlled in 65 eyes (97.01%) including 34 eyes (97.14%) and 31 eyes (96.88%) in 20G and 23G group respectively, which was not significantly different (chi;2=0.004,P＞0.05). At last follow-up, There was no statistical difference of visual acuity between the two groups (t=3.12, P＞0.05). Fourteen eyes underwent silicone oil tamponade including 13 eyes (37.14%) and 1 eye (3.13%) in 20G and 23G group respectively, which was significantly different (chi;2=11.703, P＜0.05). Nine eyes underwent reoperation (13.43%), including 8 eyes (22.86%) and 1 eye (3.13%) in 20G and 23G group respectively, which was significantly different (chi;2=5.597,P＜0.05). The 8 re-operated eyes in 20G group included 1 eye of recurrent endophthalmitis and 7 eyes with retinal detachment, the 1 re-operated eye in 23G group was of recurrent endophthalmitis. There was significantly different (chi;2=7.147,P＜0.05) for the rate of retinal detachment between the 2 groups. There were 40 eyes with bulbar conjunctiva scar including 35 eyes (100.00%) and five eyes (15.63%) in 20G and 23G group. Conclusion 23G vitrectomy is an effective treatment for infectious endophthalmitis with shorter surgery time, lower reoperation rate, lower retinal reattachment rate and fewer bulbar conjunctiva scar.