At 15 requests/user/day (totaling 4,500 queries/mo), running Translation on Microsoft Phi-3.5 MoE costs $18.225 / month ($1.822 / user / mo).
Translation for a 10 active users (alpha / internal testing) audience means 150 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 Microsoft Phi-3.5 MoE is $0.00405, of which $0.00/mo comes back from prompt caching. Per active user that's $1.822/mo — the number to hold against your pricing tier and gross margin. Charging at least $9.112/user/mo keeps gross margin at or above 80%.
Quick answer
For 10 Active Users, Translation on Microsoft Phi-3.5 MoE is modeled at 4,500 requests per month, costing $18.225 per month or $1.822 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 10 Active Users generating 4,500 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 (10 Active Users) | Cost / User / Mo | Annual Spend | Context Limit |
|---|---|---|---|---|---|
| Microsoft Phi-3.5 MoE (Current) | microsoft | $18.225 | $1.822 | $218.70 | 128,000 |
| GPT-5.6 Luna | openai | $33.592 | $3.359 | $403.11 | 1,050,000 |
| GPT-5.4 mini | openai | $125.972 | $12.597 | $1,511.663 | 256,000 |
| GPT-5.4 nano | openai | $34.83 | $3.483 | $417.96 | 128,000 |
| GPT-4o mini | openai | $17.972 | $1.797 | $215.663 | 128,000 |
| Claude Haiku 4.5 | anthropic | $143.213 | $14.321 | $1,718.55 | 1,000,000 |
At 15 queries per user/day with Microsoft Phi-3.5 MoE, the estimated cost is $1.822 per monthly active user (MAU). Total monthly bill for 10 Active Users is $18.225.
Assuming a 15% cache hit rate on repeated system and context tokens, prompt caching saves $0.00 per month ($0.00/year) on Microsoft Phi-3.5 MoE.
With a direct COGS of $1.822 per user/month on Microsoft Phi-3.5 MoE, charging at least $9.112 per user/month ensures an 80%+ software gross margin.