Driven primarily by variations in earth’s axis wobble, tilt, and orbit eccentricity, our planet experienced massive glacial/interglacial reorganizations of climate and atmospheric CO2 concentrations during the Pleistocene (2.58 Ma–11.7 ka). Even after decades of research, the underlying climate response mechanisms to these astronomical forcings have not been fully understood. To further quantify the sensitivity of the earth system to orbital-scale forcings we conducted an unprecedented quasi-continuous coupled general climate model simulation with the Community Earth System Model version 1.2 (CESM1.2, ~ 3.75° horizontal resolution), which covers the climatic history of the past 3 million years ago (3 Ma). In addition to the astronomical insolation changes, CESM1.2 is forced by estimates of CO2 and ice-sheet topography which were obtained from a simulation previously conducted with the CLIMBER-2 earth system model of intermediate complexity. Our 3 Ma simulation consists of 42 transient interglacial/glacial simulation chunks, which were partly run in parallel to save computing time. The chunks were subsequently merged, accounting for spin-up and overlap effects to yield a quasi-continuous trajectory. The computer model data were compared against a plethora of paleo-proxy data and large-scale climate reconstructions. For the period from the Mid-Pleistocene Transition (MPT, ~1 Ma) to the late Pleistocene we find good agreement between simulated and reconstructed temperatures in terms of phase and amplitude (−5.7 °C temperature difference between Last Glacial Maximum and Holocene). For the earlier part (3 Ma–1 Ma), differences in orbital-scale variability occur between model simulation and the reconstructions, indicating potential biases in the applied CO2 forcing. Our model-proxy data comparison also extends to the westerlies, which show unexpectedly large variance on precessional timescales, and hydroclimate variables in major monsoon regions. Eccentricity-modulated precessional variability is also responsible for the simulated changes in the amplitude and flavours of the El Niño-Southern Oscillation. We further identify two major modes of planetary energy transport, which played a crucial role in Pleistocene climate variability: the first obliquity and CO2-driven mode is linked to changes in the equator-to-pole temperature gradient; the second mode regulates the inter-hemispheric heat imbalance in unison with the eccentricity-modulated precession cycle. During the MPT, a pronounced qualitative shift occurs in the second mode of planetary energy transport: the post-MPT eccentricity-paced variability synchronizes with the CO2 forced signal. This synchronized feature is coherent with changes in global atmospheric and ocean circulations, which might contribute to an intensification of glacial cycle feedbacks and amplitudes. Comparison of this paleo-simulation with greenhouse warming simulations reveals that for a RCP8.5 greenhouse gas emission scenario, the projected global mean surface temperature changes over the next 7 decades would be comparable to the late Pleistocene glacial-interglacial range; but the anthropogenic warming rate will exceed any previous ones by a factor of ~100.