At 15 requests/user/day (totaling 45,000,000 queries/mo), running Translation on Cerebras — Qwen 3 32B costs $288,000.00 / month ($2.88 / user / mo).
Total expected API invoice for 100,000 Active Users generating 45,000,000 queries.
Direct inference cost per MAU to model your SaaS pricing tiers and gross margins.
Savings generated by exploiting 15% cache hits on prompt context.
| Model | Provider | Monthly Spend (100,000 Active Users) | Cost / User / Mo | Annual Spend | Context Limit |
|---|---|---|---|---|---|
| Cerebras — Qwen 3 32B (Current) | cerebras | $288,000.00 | $2.88 | $3,456,000.00 | 128,000 |
| GPT-5.6 Luna | openai | $335,925.00 | $3.359 | $4,031,100.00 | 1,050,000 |
| GPT-5.4 mini | openai | $1,259,718.75 | $12.597 | $15,116,625.00 | 256,000 |
| GPT-5.4 nano | openai | $348,300.00 | $3.483 | $4,179,600.00 | 128,000 |
| GPT-4o mini | openai | $179,718.75 | $1.797 | $2,156,625.00 | 128,000 |
| Claude Haiku 4.5 | anthropic | $1,432,125.00 | $14.321 | $17,185,500.00 | 1,000,000 |
At 15 queries per user/day with Cerebras — Qwen 3 32B, the estimated cost is $2.88 per monthly active user (MAU). Total monthly bill for 100,000 Active Users is $288,000.00.
Assuming a 15% cache hit rate on repeated system and context tokens, prompt caching saves $0.00 per month ($0.00/year) on Cerebras — Qwen 3 32B.
With a direct COGS of $2.88 per user/month on Cerebras — Qwen 3 32B, charging at least $14.40 per user/month ensures an 80%+ software gross margin.