RC planes are insanely more maneuverable and agile compared to real aircraft, this is because they are made of very light materials (Plastic, Styrofoam, etc) resulting in very high thrust to weight ratio and very powerful control surfaces.
Agreed. But this effect also goes beyond the materials being used. Weight (and volume) will scale as the length cubed, while area will scale as length squared. The result is that the wing loading will be much lower than the real aircraft, while (as you suggest) thrust loading will be higher. Both effects lead to a far more agile airplane.
Consider for example a 1/10th scale model. Even assuming that the RC model used similar materials (which it doesn't), the model airplane would have one tenth the wing loading of the actual, full scale airplane. Neglecting for the moment the effects of boundary layer transition and the higher thrust-to-weight ratio afforded by the subscale model, both the sustained and instantaneous turn rate should increase by roughly 10X in this example, due to the effects of wing loading alone.
I wonder whether the RC model uses all the control surfaces of the real J-10, especially the canards.
This particular RC model uses thrust vectoring to compensate for its lack of fully actuated control surfaces, as should be more readily evident from the video below. The canards are fixed.
In short, an RC model is a very poor indicator for the performance of the real thing.