A general model of a solar thermophotovoltaic device is discussed both for improving the efficiency of one‐band‐gap photovoltaic cells by matching the photon energy to the band gap and for concentrating diffuse radiation. First we assume ideal components to calculate theoretical maximum efficiency. It corresponds to that of a perfect selective absorber in conjunction with a Carnot‐engine ranging from 53% for 1 sun to 85% for the highest possible irradiance of 5×104suns. The improvement over an ideal one‐gap device is roughly a factor of 2. Consideration of available materials shows that any improvement in efficiency can be expected only for high irradiance of 1000 suns. The sensitive parameter is the selectivity of the absorber‐emitter. Concentration of diffuse solar radiation is not feasible. Perspectives appear not much better than for existing technologies such as photovoltaics or solar tower applications.