calculate_estimate

Estimate the cost for a given number of input and output tokens on a specific model. Useful for budgeting API calls or comparing the cost of different approaches.

Input

Parameter	Type	Required	Description
model_name	string	Yes	Model name (fuzzy matched)
input_tokens	number	Yes	Number of input tokens
output_tokens	number	Yes	Number of output tokens

Example

Request:

{
  "model_name": "claude-sonnet-4-5",
  "input_tokens": 1000,
  "output_tokens": 500
}

Response:

Cost Estimate for claude-sonnet-4-5

  Input:  1K tokens × $3.00/1M = $0.003000
  Output: 500 tokens × $15.00/1M = $0.007500
  ─────────────────────────────
  Total:  $0.0105

Tiered pricing

For models with tiered pricing (e.g. Claude Opus), the estimate automatically splits the cost at the 200K token threshold:

Request:

{
  "model_name": "claude-opus-4",
  "input_tokens": 300000,
  "output_tokens": 50000
}

Response:

Cost Estimate for claude-opus-4

  Input (base):  200K tokens × $15.00/1M = $3.00
  Input (>200K): 100K tokens × $30.00/1M = $3.00
  Output: 50K tokens × $75.00/1M = $3.75
  ─────────────────────────────
  Total:  $9.75

When token counts are at or below 200K, the standard flat-rate format is used even for tiered models.

Notes

• Token counts must be non-negative integers
• The model name is fuzzy matched, same as get_model_details
• Pricing is based on the per-token rates from the LiteLLM registry
• Models with tiered pricing automatically split costs at the 200K token boundary