In contrast to previous approaches, new reconstructions of changes in global mean surface temperature and global mean sea level (GMSL) include large variability in GMSL throughout the Pleistocene. Here we assess these reconstructions from an energy balance perspective by using them to force a global carbon cycle model in different scenarios that capture the spread in existing CO2 ${\text{CO}}_{2}$ reconstructions. Results suggests that a high CO2 ${\text{CO}}_{2}$ scenario (300–450 ppm before 2 million years ago) is most consistent with the new climate reconstructions. When radiative forcing from the high CO2 ${\text{CO}}_{2}$ scenario and land‐ice albedo derived from the new GMSL reconstruction are corrected for other forcings and feedbacks, but ignoring any state‐dependency, these results for the past 4 million years suggest an equilibrium climate sensitivity of 1.8–1.9 K which is just below the 5%–95% confidence range of recent estimates using other approaches (2.1–4.0 K). Most reconstructions of global mean sea level (GMSL) change show an increase in amplitude together with a decrease in frequency across the Mid‐Pleistocene transition around 1 million years ago. A new reconstruction now suggests that large changes of sea level occurred throughout the last 2.5 million years. Here we use both newly reconstructed temperature and sea level to force a global carbon cycle model over the past 4 million years and simulate changes in atmospheric CO2 ${\text{CO}}_{2}$ concentration. When Earth is in energy balance, temperature change reflects the sum of the main contributors of radiative forcing changes: the greenhouse effect of atmospheric CO2 ${\text{CO}}_{2}$ and Earth's reflectivity, whose main component is caused by ice sheet coverage, which can be estimated from GMSL change. We find that a high CO2 ${\text{CO}}_{2}$ scenario (300–450 ppm before 2 million years ago) delivers most consistent results. When radiative forcing is corrected for missing forcings and feedbacks, we use our results to calculate a temperature rise of 1.8–1.9° ${}^{\circ}$C for a CO2 ${\text{CO}}_{2}$ doubling, referred to as equilibrium climate sensitivity. This result is slightly lower than the range given by other recent approaches. New temperature and global mean sea level data since 4.5 Ma with large ice‐sheet variability throughout the Pleistocene force C cycle model Energy balance requires that simulated atmospheric CO2 ${\text{CO}}_{2}$ concentration before 2 Ma must be 300–450 ppm to best agree with climate data After accounting for missing forcings and feedbacks our results for the past 4 Myr suggest that equilibrium climate sensitivity is 1.8–1.9 K New temperature and global mean sea level data since 4.5 Ma with large ice‐sheet variability throughout the Pleistocene force C cycle model Energy balance requires that simulated atmospheric CO2 ${\text{CO}}_{2}$ concentration before 2 Ma must be 300–450 ppm to best agree with climate data After accounting for missing forcings and feedbacks our results for the past 4 Myr suggest that equilibrium climate sensitivity is 1.8–1.9 K