You can use the normal plot_surface
method from matplotlib. Because it needs a X and Y array, its already plotted with the right coordinates. I always find it hard to make nice looking 3D plots, so the visual aspects can certainly be improved. :)
import gdal
from mpl_toolkits.mplot3d import Axes3D
dem = gdal.Open('gmted_small.tif')
gt = dem.GetGeoTransform()
dem = dem.ReadAsArray()
fig, ax = plt.subplots(figsize=(16,8), subplot_kw={'projection': '3d'})
xres = gt[1]
yres = gt[5]
X = np.arange(gt[0], gt[0] + dem.shape[1]*xres, xres)
Y = np.arange(gt[3], gt[3] + dem.shape[0]*yres, yres)
X, Y = np.meshgrid(X, Y)
surf = ax.plot_surface(X,Y,dem, rstride=1, cstride=1, cmap=plt.cm.RdYlBu_r, vmin=0, vmax=4000, linewidth=0, antialiased=True)
ax.set_zlim(0, 60000) # to make it stand out less
ax.view_init(60,-105)
fig.colorbar(surf, shrink=0.4, aspect=20)
