Benchmark

AI Startup Cost Benchmark: 10k MAU Infrastructure

How much does it cost to run an AI-powered SaaS with 10,000 monthly active users? CostMyAI benchmarks AI startup infrastructure costs across GPT-4o, GPT-4o mini, Claude 3.5 Haiku, Gemini 2.0 Flash, and DeepSeek V3 for a representative chatbot workload.

Benchmark assumptions

  • 10,000 monthly active users, 20 AI interactions per user per month
  • Average interaction: 400 input tokens + 200 output tokens per request
  • 200,000 total requests per month (2.4 million requests per year)
  • No batch processing discount applied (synchronous, user-facing requests)

Monthly API cost by model

  • GPT-4.1 Nano ($0.10/$0.40 per 1M): approximately $164/mo
  • Gemini 2.0 Flash ($0.10/$0.40 per 1M): approximately $164/mo
  • GPT-4o mini ($0.15/$0.60 per 1M): approximately $246/mo
  • Claude 3.5 Haiku ($0.80/$4.00 per 1M): approximately $960/mo
  • GPT-4o ($2.50/$10.00 per 1M): approximately $3,000/mo
  • Claude 3.5 Sonnet ($3.00/$15.00 per 1M): approximately $3,600/mo

The 22x cost difference between the cheapest and most expensive model for this workload demonstrates why model selection is the single most impactful cost lever for early-stage AI products.

Benchmark my real usage