Subseasonal Arctic sea ice prediction is highly needed for practical services including icebreakers and commercial ships, while limited by the capability of climate models. A bias correction methodology in this study was proposed and performed on raw products from two climate models, the First Institute Oceanography Earth System Model (FIOESM) and the National Centers for Environmental Prediction (NCEP) Climate Forecast System (CFS), to improve 60 days predictions for Arctic sea ice. Both models were initialized on July 1, August 1, and September 1 in 2018. A 60-day forecast was conducted as a part of the official sea ice service, especially for the ninth Chinese National Arctic Research Expedition (CHINARE) and the China Ocean Shipping (Group) Company (COSCO) Northeast Passage voyages during the summer of 2018. The results indicated that raw products from FIOESM underestimated sea ice concentration (SIC) overall, with a mean bias of SIC up to 30%. Bias correction resulted in a 27% improvement in the Root Mean Square Error (RMSE) of SIC and a 10% improvement in the Integrated Ice Edge Error (IIEE) of sea ice edge (SIE). For the CFS, the SIE overestimation in the marginal ice zone was the dominant features of raw products. Bias correction provided a 7% reduction in the RMSE of SIC and a 17% reduction in the IIEE of SIE. In terms of sea ice extent, FIOESM projected a reasonable minimum time and amount in mid-September; however, CFS failed to project both. Additional comparison with subseasonal to seasonal (S2S) models suggested that the bias correction methodology used in this study was more effective when predictions had larger biases.