Save 3d plot as vector format

The Sage 3D plots don't support output in a 2D vector format (like PDF), at least not for now. You can do it by calling matplotlib directly: PDF version of the image below

from mpl_toolkits.mplot3d import axes3d
from mpl_toolkits.mplot3d.art3d import Poly3DCollection
import matplotlib.pyplot as plt
import numpy as np
from matplotlib import rc
rc('text', usetex=True)

rays = [(-3, -2, 4), (0, 1, 0), (1, 0, 0), (2, 1, -4)]
extrapoints = [(-1, -1, 2), (1, 1, -2)]
cones = [[0,1,2],[0,1,3],[0,2,3],[1,2,3]]

poly3d = []
for cone in cones:
    verts = [ rays[i] for i in cone ]
    poly3d.append(verts)

polygons = Poly3DCollection(poly3d, facecolor='green', 
                            linewidths=2,
                            alpha=0.1)
polygons.set_alpha(0.3)

fig = plt.figure()
fig.set_figwidth(13.0/2.54)
fig.set_figheight(9.0/2.54)

ax = axes3d.Axes3D(fig)
ax.scatter3D(*zip(*rays), s=50, c='red')
ax.scatter3D(*zip(*extrapoints), s=50, c='blue')
ax.scatter3D([0],[0],[0], s=20, c='black')

ax.add_collection3d(polygons)  
ax.set_xlabel(r'$X$')
ax.set_ylabel(r'$Y$')
ax.set_zlabel(r'$Z$')
ax.set_xlim3d(-3, 2)
ax.set_ylim3d(-2, 1)
ax.set_zlim3d(-4, 4)

r = rays[0]
ax.text(r[0]-0.2,r[1],r[2], r'${}^{'+str(r)+'}$', 
        horizontalalignment='right',
        verticalalignment='top',
        color='red')

r = rays[3]
ax.text(r[0]-0.2,r[1],r[2], r'${}^{'+str(r)+'}$', 
        horizontalalignment='right',
        verticalalignment='top',
        color='red')

r = rays[1]
ax.text(r[0],r[1],r[2]+0.3, r'${}^{'+str(r)+'}$', 
        horizontalalignment='center',
        verticalalignment='bottom',
        color='red')

r = rays[2]
ax.text(r[0],r[1],r[2]-0.3, r'${}^{'+str(r)+'}$', 
        horizontalalignment='center',
        verticalalignment='top',
        color='red')

r = extrapoints[0]
ax.text(r[0],r[1],r[2]-0.3, r'${}^{'+str(r)+'}$', 
        horizontalalignment='center',
        verticalalignment='top',
        color='blue')

r = extrapoints[1]
ax.text(r[0]+0.2,r[1],r[2], r'${}^{'+str(r)+'}$', 
        horizontalalignment='left',
        verticalalignment='bottom',
        color='blue')


ax.text(-0.2,0,0, r'${}^0$', 
        horizontalalignment='right',
        verticalalignment='top',
        color='black')

# plt.show()

for axis in [ax.w_xaxis, ax.w_yaxis, ax.w_zaxis]:
    axis.get_major_locator()._integer = True
    for t in axis.get_ticklabels(): 
        t.set_fontsize(6)

plt.savefig('fig_NablaBsing.pdf', bbox_inches='tight')