Langasite (LGS) is a promising material for SAW applications at high temperature. However, the temperature coefficients of LGS material constants are not accurate enough to perform reliable simulations, and therefore to make good use of available design tools, above 300°C. In the first part of the paper, we describe a new possible way to derive these coefficients in a wider temperature range. The method is based on Simulated Annealing, a well-known optimization algorithm. The algorithm converges toward a set of optimized temperature coefficients of the stiffness constants which are used to perform accurate simulations up to at least 800°C. In the second part, a deeper analysis of the algorithm outputs demonstrates some of its strengths but also some of its main limitations. Possible solutions are described to predict and then improve the accuracy of the optimized coefficient values. In particular, one solution making use of additional BAW target curves is tested. A promising solution to extend the optimization to the temperature coefficients of piezoelectric constants is also discussed.