The most powerful explosions on the Sun-in the form of bright flares, intense storms of solar energetic particles (SEPs), and fast coronal mass ejections (CMEs)-drive the most severe space-weather storms. Proxy records of flare energies based on SEPs in principle may offer the longest time base to study infrequent large events. We conclude that one suggested proxy, nitrate concentrations in polar ice cores, does not map reliably to SEP events. Concentrations of select radionuclides measured in natural archives may prove useful in extending the time interval of direct observations up to ten millennia, but as their calibration to solar flare fluences depends on multiple poorly known properties and processes, these proxies cannot presently be used to help determine the flare energy frequency distribution. Being thus limited to the use of direct flare observations, we evaluate the probabilities of large-energy solar events by combining solar flare observations with an ensemble of stellar flare observations. We conclude that solar flare energies form a relatively smooth distribution from small events to large flares, while flares on magnetically active, young Sun-like stars have energies and frequencies markedly in excess of strong solar flares, even after an empirical scaling with the mean coronal activity level of these stars. In order to empirically quantify the frequency of uncommonly large solar flares extensive surveys of stars of near-solar age need to be obtained, such as is feasible with the Kepler satellite. Because the likelihood of flares larger than approximately X30 remains empirically unconstrained, we present indirect arguments, based on records of sunspots and on statistical arguments, that solar flares in the past four centuries have likely not substantially exceeded the level of the largest flares observed in the space era, and that there is at most about a 10% chance of a flare larger than about X30 in the next 30years.