Monitoring Performance and Costs

AI is expensive if unmanaged.

The Formula

Cost = (Input Tokens * Price) + (Output Tokens * Price).

• Images: Fixed price per image.
• Video: Price per second.

Estimation

Before launch, calculate the Unit Economics.

• Avg Input: 500 tokens.
• Avg Output: 200 tokens.
• Task: "Summarize Email".
• Model: Gemini Flash ($0.35 / 1M input).
• Cost per Task: $0.000175. (Very cheap).

Optimization

1. Use Flash: 10x-20x cheaper than Pro.
2. Shorten Prompts: Remove polite filler ("Please be so kind as to...").
3. Limit Output: Use max_output_tokens param to prevent the model from writing a novel when you wanted a sentence.

Summary

If you track tokens, you control costs.

Module 11 Complete! You are ready for Production. In the final step, we build the Capstone Project.