An analog-based air quality forecast bias-correction technique, developed through years of collaboration between NOAA/PSL, NCAR, and NCEP, still has some room for improvement, especially for random high pollution events like wild fires. Generating reliable forecasts of extreme values of ozone and PM2.5 is an important aspect of the CMAQ post-processing system. Several recent changes to the post-processing code, including approaches for fire-associated PM2.5 predictions include: 1) a short training period approach for cases of high PM2.5 related to intense wild fires is implemented and evaluated on several examples including extreme fire event in the western states of Washington, Oregon and California in September, 2020. Several possible training scenarios are discussed. Ready for deployment, but has only been demonstrated for 1 year of data. 2) developing a new more flexible code that allows for the use of input observational data in new CSV format which have the advantages of more monitors for both ozone and PM2.5, and that the file refresh rate is much faster. Now deployed in both CMAQ versions, AQMv6, current operations, and AQMv7 pre-operational. 3) filling data gaps of 1-2 hours in the observations, which helps to avoid air-quality mapping with artificial bull-eyes high value ozone or PM2.5 cases. Delivered 2023 October 27 as a patch release. Now deployed in AQMv7 pre-operational for final testing. All these improvements are evaluated with the goal of implementing them in the next generation air quality (AQ) prediction system for the United States being developed at NCEP.