Cost Estimation for Report Summarization Using Google Gemini Pro

Scenario Assumptions:

• Input: 10-page PDF containing mostly numerical tables.
• Output: Summarized form with commentary, around 3-4 pages.
• Context Size Considerations:
  • 1K tokens ≈ 750 words.
  • A full-page of dense tabular data likely contains 500-750 words.
  • So, 10 pages of tabular data ≈ 5,000 - 7,500 words.
  • Converted to tokens: 6,700 - 10,000 tokens.
  • The summarization output (3-4 pages) ≈ 1,500 - 3,000 words ≈ 2,000 - 4,000 tokens.
  • Total token usage per summarization query:
    • Input Tokens: ~ 8,000 tokens.
    • Output Tokens: ~ 3,000 tokens.
    • Total Tokens per query: 11,000 tokens.

---

Cost Calculation:

From the image, the cost for Google Gemini Pro:

• Input Token Cost: $0.000125 per 1K tokens.
• Output Token Cost: $0.000375 per 1K tokens.

Cost per summarization run:

• Input Token Cost:
  • (8,000 / 1,000) × 0.000125 = $0.001
• Output Token Cost:
  • (3,000 / 1,000) × 0.000375 = $0.001125
• Total cost per summarization: $0.002125 per run

---

Monthly Cost (Running Once per Day)

• Runs per month: 30
• Total monthly cost:
  • 30 × 0.002125 = $0.06375 ≈ 6.4 cents

This is a very low cost (~$0.06 per month) because the number of tokens used per query is not excessive compared to large-scale chatbot use cases.

---

Scaling Considerations

If scaled up for 100 reports per day, the cost would be:

• Daily Cost:
  • 100 × 0.002125 = $0.2125
• Monthly Cost:
  • 0.2125 × 30 = $6.38

So, running 100 reports per day for a month would cost $6.38 per month.

---

Final Summary

• Single report summarization (10 input pages, 3-4 output pages) costs ~ $0.0021 per run.
• Running this once per day for a month costs ~ $0.064 (6.4 cents).
• Scaling up to 100 reports per day costs ~ $6.38 per month.

This is a cost-effective implementation for automated LLM-based summarization. Would you like to tweak these numbers for different use cases or add additional complexity like multi-document processing?