A long-term sublimation model to explain how Phaethon could provide the Geminid stream is proposed. We find that it would take ∼ 6 Myr or more for Phaethon to lose all of its internal ice (if ever there was) in its present orbit. Thus, if the asteroid moved from the region of a 5:2 or 8:3 mean motion resonance with Jupiter to its present orbit less than 1 Myr ago, it may have retained much of its primordial ice. The dust mantle on the sublimating body should have a thickness of at least 15 m but the mantle could have been less than 1 m thick 1000 years ago. We find that the total gas production rate could have been as large as 10 27 s −1 then, and the gas flow could have been capable of lifting dust particles of up to a few centimeters in size. Therefore, gas production during the past millennium could have been sufficient to blow away enough dust particles to explain the entire Geminid stream. For present-day Phaethon, the gas production is comparatively weak. But strong transient gas release with a rate of ∼ 4.5 × 10 19 m −2 s −1 is expected for its south polar region when Phaethon moves from 0 • to 2 • mean anomaly near perihelion. Consequently, dust particles with radii of <∼ 260 µm can be blown away to form a dust tail. In addition, we find that the large surface temperature variation of > 600 K near perihelion can generate sufficiently large thermal stress to cause fracture of rocks or boulders and provide an efficient mechanism to produce dust particles on the surface. The time scale for this process should be several times longer than the seasonal thermal cycle, thereby dominating the cycle of appearance of the dust tail.