Background & Aims: Non-alcoholic fatty liver disease (NAFLD) patients with diabetes are at increased risk of cirrhosis and liver-related death, and thus accurate fibrosis assessment in these patients is important. We examined the ability of non-invasive fibrosis models to determine cirrhosis and outcomes in NAFLD patients with and without diabetes. Methods: Non-alcoholic fatty liver disease patients diagnosed between 2006 and 2015 had Hepascore, NAFLD fibrosis score (NFS), APRI and FIB-4 scores calculated at baseline and were followed up for outcomes of overall and liver-related mortality/liver transplantation, hepatic decompensation and hepatocellular carcinoma (HCC). Model accuracy was determined by Harrell's C-index and by Kaplan-Meier analysis. Results: A total of 284 patients (53% diabetic, 15% cirrhotic) were followed up for a median of 51.4 months, (range 6.1-146). During follow-up, diabetic patients had a greater risk of liver-related death/transplantation, HR 3.4 (95% CI 1.2-9.1) decompensation, HR 4.7 (95% CI 2.0-11.3) and HCC, HR 2.9 (95% CI 1.2-7.3). Among 241 subjects with a baseline liver biopsy, the accuracy of Hepascore, APRI and FIB-4 for predicting cirrhosis was lower amongst diabetics compared to non-diabetics (P <.005 for all). Model accuracy apart from Hepascore, was also significantly lower for predicting liver death/transplantation in patients with diabetes. No patient with a low fibrosis score and without diabetes developed liver decompensation or HCC, whereas up to 21% of diabetic patients with a low fibrosis score developed liver decompensation and up to 27% developed HCC at 5 years. Conclusions: Non-invasive scoring systems are less accurate at predicting cirrhosis and liver-related outcomes in patients with NAFLD and diabetes.