If all my variables' first difference is stationary, does it mean my variables are all stationary? I wonder all my variables' first difference is stationary, does it mean my variables are all stationary in time series? What model should I use then? what the format would be? How to address this question in Stata?
 A: I wonder all my variables' first difference is stationary, does it mean my variables are all stationary in time series?
No. For example, if $y_t = \beta y_{t-1} + \varepsilon_{t}$, where $|\beta| \ge 1$, and $\varepsilon \sim \mathcal{N}(\mu,\sigma)$ then $y$ is a non-stationary process, however $\Delta y_{t} = y_t - y_{t-1}$ (the first difference of $y$) is stationary, because $\Delta y_{t} = \varepsilon_{t}$, and $\varepsilon$ is strongly stationary. (If $0 < |\beta| < 1$ in the simple model here, $y$ is said to be 'weakly stationary'.)
What model should I use then?
This depends on what your research question is. In general, if one can find stationary differences of time series data, one has options for modeling and inference which will not run afoul of spurious correlation. (Albeit, selection bias, and confounding via backdoor paths may still be involved.)
what would the format would be?
Not sure what you are asking here? The data would generally be organized in a unit-period format.
How to address this question in Stata?
Depends on what you want to do. Maybe read the help files for tsset, xtset, xtunitroot, xtmixed, and tsvarlist? One can certainly do a lot with time series models in Stata.
