The self-accelerating braneworld model (DGP) seems to provide a simple alternative to the the standard $\Lambda$CDM cosmology to explain the current cosmic acceleration, which is strongly indicated by measurements of Type Ia supernovae, as well as other concordant observations. In this work, we investigate observational constraints on this scenario from gravitational lensing statistics using the Cosmic Lens All-Sky Survey (CLASS) lensing sample. We show that a large parameter space of the DGP model is in good agreement with this radio source gravitational lensing sample. In the flat case, $\Omega_\mathrm{K}=0$, the likelihood is maximized, ${\cal L}={\cal L_\mathrm{max}}$, for $\1 = 0.30_{-0.11}^{+0.19}$. If we relax the prior on $\Omega_\mathrm{K}$, the likelihood peaks at $\{\1,\2 \} \simeq \{0.29, 0.12\}$, just slightly in the region of open models. However the confidence contours are pretty elongated so that we can not discard either close or flat or open models