What a Single Call Actually Costs You in USDC: A Per-Request Fee Worksheet for Hobbyists

Why compute per-call cost?

When you're a hobbyist running small experiments, every cent counts. Knowing the exact USDC deducted per call helps you budget and compare models. This worksheet walks you through hand-calculating the cost for one prompt-response pair on each of the six models available through our API aggregator: Claude, GPT, DeepSeek, Qwen, GLM, and Kimi.

The cost formula

For any model, the USDC deducted per call is:

cost = (inputtokens * inputpricepertoken + outputtokens * outputpricepertoken) * 1.3

• input_tokens: number of tokens in your prompt (including any system message).
• output_tokens: number of tokens in the model's response.
• inputpricepertoken and outputpricepertoken: official prices from the model provider (in USDC per token).
• 1.3: the multiplier applied by our service.

Note: All prices are in USDC. Official prices vary by model and are subject to change—refer to the provider's pricing page for current rates.

Step-by-step worksheet

Follow these steps for each model:

1. Tokenize your prompt – Use a tokenizer tool or the model's API to count input tokens. For rough estimates, remember that 1 token ≈ 4 characters in English.
2. Get the response – Send your prompt and record the output token count from the API response (e.g., usage.output_tokens).
3. Look up official prices – Find the model's per-token input and output prices on the provider's website.
4. Apply the formula – Multiply tokens by prices, sum, then multiply by 1.3.
5. Round to the nearest USDC micro-unit – Most systems deduct to 6 decimal places (micro-USDC).

Example calculation (hypothetical)

Suppose you send a 100-token prompt to Model A and get a 200-token response. If Model A's official prices are $0.000001 per input token and $0.000002 per output token:

• Input cost: 100 * 0.000001 = 0.0001 USDC
• Output cost: 200 * 0.000002 = 0.0004 USDC
• Subtotal: 0.0005 USDC
• With 1.3x multiplier: 0.0005 * 1.3 = 0.00065 USDC

So you'd be charged 0.00065 USDC for that call.

Model-specific notes

Each model has different pricing structures, tokenization rules, and context limits. Here are key considerations when computing costs:

• Claude: Anthropic uses a distinct tokenizer; input and output prices differ. Long prompts may hit context limits, so token count matters.
• GPT: OpenAI's models often have separate prices for different variants (e.g., GPT-4 vs GPT-3.5). Always use the exact model name's pricing.
• DeepSeek: Known for competitive pricing; check if the model charges differently for input vs output.
• Qwen: Alibaba's models may have tiered pricing based on context length. Verify the price for your specific request size.
• GLM: Zhipu AI's models sometimes offer different rates for different versions (e.g., GLM-4 vs GLM-3).
• Kimi: Moonshot AI's models may have promotional rates or token-based tiers.

Accounting for token splits

Most APIs report usage as separate input and output token counts. Always use both—never assume a flat rate. For example, if a model charges $0.000001 per input token and $0.000002 per output token, a call with 500 input and 1000 output tokens costs:

(500 * 0.000001 + 1000 * 0.000002) * 1.3 = (0.0005 + 0.002) * 1.3 = 0.0025 * 1.3 = 0.00325 USDC

If you only used the input price, you'd underestimate the cost.

Checking your USDC deduction

After making a call, you can verify the deduction in your account dashboard. The amount should match your hand calculation (within rounding). If it doesn't, double-check:

• Token counts (input and output)
• Official prices (ensure you're using the correct model version)
• The 1.3 multiplier

Reducing per-call cost

While the 1.3x multiplier is fixed, you can lower your USDC spend by:

• Shortening prompts (fewer input tokens)
• Requesting concise responses (fewer output tokens)
• Choosing models with lower official prices for simple tasks
• Batching multiple queries into one call when possible

Contributor credit

If you contribute a key that gets used, you earn credit at official price × 1.1 (or × 1.2 for premium keys) in USDC. This is separate from the cost you pay as a user; it's a reward for sharing your API access.