At 15 requests/user/day (totaling 22,500 queries/mo), running Translation on Qwen 3 32B Instruct costs $137.925 / month ($2.759 / user / mo).
Translation for a 50 active users (private beta) audience means 750 requests per day on average, each carrying 5,000 input and 5,500 output tokens at a 15% cacheable share — so the effective per-request cost on Qwen 3 32B Instruct is $0.00613, of which $6.075/mo comes back from prompt caching. Per active user that's $2.759/mo — the number to hold against your pricing tier and gross margin. Charging at least $13.793/user/mo keeps gross margin at or above 80%.
Quick answer
For 50 Active Users, Translation on Qwen 3 32B Instruct is modeled at 22,500 requests per month, costing $137.925 per month or $2.759 per active user.
Method & trust
This projection combines the selected active-user tier, requests per user per day, input/output tokens per request, and the modeled cache share. It is an operating estimate, not a provider invoice; production traffic and retries should be measured against it.
Total expected API invoice for 50 Active Users generating 22,500 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 (50 Active Users) | Cost / User / Mo | Annual Spend | Context Limit |
|---|---|---|---|---|---|
| Qwen 3 32B Instruct (Current) | qwen | $137.925 | $2.759 | $1,655.10 | 128,000 |
| GPT-5.6 Terra | openai | $1,679.625 | $33.593 | $20,155.50 | 1,050,000 |
| GPT-5.4 Workhorse | openai | $2,099.531 | $41.991 | $25,194.375 | 256,000 |
| o3-mini | openai | $658.969 | $13.179 | $7,907.625 | 200,000 |
| o4-mini | openai | $654.328 | $13.087 | $7,851.938 | 256,000 |
| o1-mini | openai | $658.969 | $13.179 | $7,907.625 | 128,000 |
At 15 queries per user/day with Qwen 3 32B Instruct, the estimated cost is $2.759 per monthly active user (MAU). Total monthly bill for 50 Active Users is $137.925.
Assuming a 15% cache hit rate on repeated system and context tokens, prompt caching saves $6.075 per month ($72.90/year) on Qwen 3 32B Instruct.
With a direct COGS of $2.759 per user/month on Qwen 3 32B Instruct, charging at least $13.793 per user/month ensures an 80%+ software gross margin.