The radio emission of flares at wavelengths from millimeter to decameter waves includes a large variety of emission processes. They can be considered as different diagnostic tools particularly suited for the analysis of non-thermal electron distributions, enhanced levels of various kinds of plasma, waves and plasma phenomena. Incoherent gyrosynchrotron emission at millimeter and centimeter waves provides higher sensitivity for observing MeV electrons than existing hard X-ray (HXR) and gamma-ray satellites. Very intense coherent emissions are observed at wavelengths longer than about 10 cm, weaker ones from about 4 cm. They are caused by plasma instabilities driving various wave modes that in turn may emit observable radio waves. Particularly important are type III bursts, caused by electron beams exciting Langmuir waves. Their trace in the corona points back to the acceleration region of the electrons. Less known are radio emissions from trapped electrons driving loss-cone unstable waves. This is the interpretation usually given to decimetric type IV emission. These types of coherent radio emission give clues on the geometry and plasma parameters near the acceleration region. More speculative are emissions that are directly produced by the acceleration process. A possible group of such phenomena are narrowband, short peaks of emission. Narrowband spikes are seen sometimes at frequencies above the start of metric type III events. There is mounting evidence for the hypothesis that these spikes coincide with the energy release region. Much less clear and highly controversial is the situation for decimetric spikes, which are associated with HXR flares. More frequently than spikes, however, there is fluctuating broadband decimetric emission during the HXR phase of flares. The use of these coherent radio emissions as a diagnostic tool for the primary energy release requires a solid understanding of the emission process. At the moment we are still far away from an accepted theory. Only careful comparisons with complementary observations of energetic electrons and the thermal coronal background in EUV lines and soft X-rays can put coherent emissions into context and test the different scenarios. The comparison with HXR, millimeter and centimeter observations will be necessary to derive quantitative results on energy release. In combination with other wavelengths and their recent imaging capabilities, exciting new possibilities are now opening for radio diagnostics.