We discussed this a bit here: Another API in the competition
While Gemini has to be the core focus of your work, you’re certainly allowed to use other models where they make sense.
Your reasoning is a bit odd, however. There is no specific “GPU quota” for Gemini, although there is a limit on the number of tokens per minute or day. Those should reset, however.
Use the best model for the project. Remember that you’re submitting code as well for them to review.