Every time you use Apers, your request runs through several AI processes: parsing your query, reasoning, retrieving data, and generating responses. To make this transparent, we built Smart Request Credit, a system that calculates costs precisely based on the true compute used.
This guide explains exactly how that calculation works so you always know what you’re paying for.
1. The Two-Stage Charge
Each request goes through two stages:
- Prepare Stage: When you submit a request, Apers charges 1 credit upfront. This covers the base compute cost for initializing the request.
- Final Stage: Once the AI completes the task, Apers calculates the real compute cost using token-level data, then deducts additional credits based on that cost.
2. The Credit Formula
The total number of credits is determined by this formula:
- The 1 represents the minimum initializing request cost.
- The floor function means fractional results are always rounded down, never up.
- The 0.10 denominator defines the credit unit
3. Token-Based Unit Costs
Token is the basic unit of Language Model-based AI system. You can learn more by checking this article. Each token (a few characters of text) contributes differently to compute cost, depending on what the model is doing.
We also optimize Apers system to save users costs by implementing Cache Read/Write. These rates ensure large and repetitive requests are still efficient, i.e., cached data dramatically reducing cost on subsequent runs.
4. How the Floor Function Affects You
The floor() in the formula makes your charges stable and predictable.
For example, if your true compute cost is $0.099, the system charges floor(0.099 ÷ 0.10) = 0, so you only pay the 1 upfront credit.
Only when the total exceeds $0.10 does another credit get added.
This means you never overpay for fractional usage.
5. Real Examples
Let’s look at how Smart Request Credit works in different scenarios.
Example 1: Simple Query
Prompt: “Explain what is cap rate.”
Approximate tokens: 50 input + 150 output = 200 tokens
Cost = (50 × 0.000006) + (150 × 0.00003) = 0.0048
Credits = 1 + floor(0.0048 ÷ 0.10) = 1 credit total
→ You’re charged 1 credit for a short answer.
Example 2: Moderate Query
Prompt: “Read five documents and compare the differences.”
Approximate tokens: 6,000 input + 2,000 output = 8,000 tokens
Cost = (6,000 × 0.000006) + (2,000 × 0.00003) = 0.096
Credits = 1 + floor(0.096 ÷ 0.10) = 1 credit total
→ If the documents are slightly longer (≈9,000 tokens total), the cost passes 0.10 → 2 credits.
Example 3: Complex Task
Prompt: “Plan and build a 10-tab Excel spreadsheet with linked cash flow and sensitivity analysis.”
Approximate tokens: 20,000 input + 10,000 output = 30,000 tokens
Cost = (20,000 × 0.000006) + (10,000 × 0.00003) = 0.42
Credits = 1 + floor(0.42 ÷ 0.10) = 5 credits total
→ You’re charged 5 credits for a multi-step, high-compute task.
6. Summary
7. What This Means for You
- Every request always costs at least 1 credit.
- Additional credits depend entirely on actual compute.
- Thanks to floor rounding, you only pay for complete increments of usage.
- Because compute is metered directly, Apers never limits or swaps models. You always access the most powerful AI, every time.
Smart Request Credit gives you clarity, fairness, and full access to the highest model performance, whether you’re asking a quick question or orchestrating complex workflows.
