To investigate the thermal smoke characteristics in large spaces of high-rise buildings, a series of full-scale fire experiments are conducted under different heat release rates (HRR). The experiments were conducted in a three-story building. A cross-shaped thermocouple array was deployed along the fire source centreline, beneath the ceiling, and in adjacent spaces to systematically measure fire plume temperatures, ceiling jet temperature distributions, and horizontal smoke spread patterns. The experimental results indicate that the centerline temperature of the fire plume decreases with increasing height. As the HRR increases, the temperature gradient diminishes. However, under high HRR conditions (≥ 1250 kW), incomplete combustion near the fire source at 0.3 m leads to an anomalous temperature drop. Due to the restricted ceiling height, the fire plume failed to develop a distinct buoyant plume region, thus necessitating modifications to the McCaffrey model. Furthermore, the classical Alpert and Heskestad ceiling jet models show significant deviations from the experimental data obtained under these strong plume impingement conditions. This paper presents a modified dimensionless temperature distribution model based on experimental data. The smoke spread velocity is influenced by the HRR. However, under high HRR conditions (≥ 1000 kW), excessive smoke volume can hinder initial dispersion. The findings of this study can provide experimental data and theoretical references for fire risk assessment and fire protection design in large spaces of high-rise buildings.