Gemini 1.5 Flash is quite cheap. It doesn’t matter. And we can implement preprocessing to exclude unnecessary lines of logs with heuristics before sending a request to Gemini. That enables us to reduce the cost and response time.

https://cloud.google.com/vertex-ai/generative-ai/pricing