Rather than having many different models generate their own code, it would be much better to choose a base model, examine the code generated by others' base models, approve their optimization recommendations, and move on. In this way, it is more confident about the code and it is more possible to reach the optimal value with the consortium decision.