Scalable multi-robot transition is essential for ubiquitous adoption of robots. As a step towards it, a computationally efficient decentralized algorithm for continuous-time trajectory optimization in multi-robot scenarios based upon model predictive control is introduced. The robots communicate only their current states and goals rather than sharing their whole trajectory; using this data each robot predicts a continuous-time trajectory for every other robot exploiting optimal control based motion primitives that are corrected for spatial inter-robot interactions using least squares. A non linear program (NLP) is formulated for collision avoidance with the predicted trajectories of other robots. The NLP is condensed by using time as a parametrization resulting in an unconstrained optimization problem and can be solved in a fast and efficient manner. Additionally, the algorithm resizes the robot to accommodate it's trajectory tracking error. The algorithm was tested in simulations on Gazebo with aerial robots. Early results indicate that the proposed algorithm is efficient for upto forty homogeneous robots and twenty one heterogeneous robots occupying 20\% of the available space.