To ensure proper process control of a flash furnace precise real-time information of various operational parameters such the slag chemistry is required. Currently the most used methods to obtain this information are based on X-ray diffraction (XRD). Although XRD techniques are highly accurate, they provide information every couple of hours which hinders their applicability as real-time measurement instruments that aid process control. In this work, an optical probe system supported by a deep learning (DL) model is implemented and validated to monitor in real time the slag copper content in direct-to-blister flash-smelting process. In particular, the optical probe system employs Planck’s radiometry to calculate the slag surface irradiance, temperature, and spectral emissivity and then combines this radiation information with the furnace copper concentrate feed rate information to input DL models. The proposed system was validated in a real-case usage scenario in the Olympic DAM’s direct-to-blister flash furnace over November 2024. The results show that the proposed optical probe can provide continuous slag surface temperature and copper content estimation with high accuracy, while operating without interruptions in a hostile high-temperature environment.