In this work, we numerically calculate the thermal radiation efficiency of the baryonic outflow. The possible outflow acceleration in the transparent stage, which lowers thermal radiation efficiency, has been taken into account. In the standard internal shock model for the prompt emission, the fast shells should move with a typical Lorentz factor ∼5 Γi otherwise the γ‐ray burst (GRB) efficiency will be in disagreement with the observations, where Γi is the bulk Lorentz factor of the shocked/emitting region. The photosphere radius of these fast shells is small and the thermal radiation is too strong to be effectively outshone by the internal shock emission. This is particularly the case for some extremely bright events having Γi∼ 103, like GRBs 080319B and 080916C. The absence of a distinct thermal component in the spectrum of most GRBs challenges the standard internal shock model and may suggest a non‐baryonic (magnetic) outflow component. Though the magnetic outflow model seems favoured by more and more data, it can hardly reproduce the typical GRB spectrum. In the photosphere‐gradual magnetic dissipation scenario, the spectrum cuts off at ∼1 GeV, too low to account for the observations of GRBs 080916C. In the sudden magnetic energy dissipation model, the low‐energy spectrum is expected to be Fν∝ν−1/2, too soft to be consistent with the data Fν∝ν0. We speculate that the low‐energy spectrum puzzle could be unveiled by the mechanism that particles, in the magnetic dissipation process, are repeatedly accelerated.