# cost_analysis.py
from dataclasses import dataclass
[docs]
@dataclass
class ModelPricing:
"""Pricing structure for OpenAI models per 1M tokens."""
input_price: float # Price per 1M input tokens (uncached)
cached_input_price: float # Price per 1M cached input tokens
output_price: float # Price per 1M output tokens
# Pricing registry for different OpenAI models
# Prices are per 1,000,000 tokens
MODEL_PRICING = {
"gpt-4o": ModelPricing(input_price=2.50, cached_input_price=1.25, output_price=10.00),
"gpt-4o-2024-11-20": ModelPricing(input_price=2.50, cached_input_price=1.25, output_price=10.00),
"gpt-4o-2024-08-06": ModelPricing(input_price=2.50, cached_input_price=1.25, output_price=10.00),
"gpt-4o-2024-05-13": ModelPricing(input_price=5.00, cached_input_price=2.50, output_price=15.00),
"gpt-4o-mini": ModelPricing(input_price=0.150, cached_input_price=0.075, output_price=0.600),
"gpt-4o-mini-2024-07-18": ModelPricing(input_price=0.150, cached_input_price=0.075, output_price=0.600),
"o1-preview": ModelPricing(input_price=15.00, cached_input_price=7.50, output_price=60.00),
"o1-preview-2024-09-12": ModelPricing(input_price=15.00, cached_input_price=7.50, output_price=60.00),
"o1-mini": ModelPricing(input_price=3.00, cached_input_price=1.50, output_price=12.00),
"o1-mini-2024-09-12": ModelPricing(input_price=3.00, cached_input_price=1.50, output_price=12.00),
}
[docs]
def get_model_pricing(model: str) -> ModelPricing:
"""
Get pricing for a specific model.
Args:
model: The model name (e.g., "gpt-4o", "gpt-4o-mini")
Returns:
ModelPricing object with pricing information
Raises:
ValueError: If model is not found in the pricing registry
"""
if model not in MODEL_PRICING:
raise ValueError(
f"Model '{model}' not found in pricing registry. Available models: {', '.join(MODEL_PRICING.keys())}"
)
return MODEL_PRICING[model]
[docs]
def analyze_cost(usage_list, model: str = "gpt-4o-mini"):
"""
Analyze and compute the total cost based on cumulative usage details.
Each usage object is expected to have:
- "prompt_tokens": total prompt tokens used,
- "completion_tokens": total output tokens,
- "prompt_tokens_details": an object containing "cached_tokens".
Args:
usage_list: List of usage objects from OpenAI API responses
model: The OpenAI model name for pricing lookup (default: "gpt-4o-mini")
Returns:
Dictionary with token counts and computed costs including:
- total_cached_tokens: Total cached prompt tokens
- total_prompt_tokens: Total prompt tokens (cached + uncached)
- total_output_tokens: Total completion/output tokens
- total_uncached_tokens: Total uncached prompt tokens
- cost_uncached: Cost of uncached tokens
- cost_cached: Cost of cached tokens
- cost_output: Cost of output tokens
- total_cost: Total estimated cost
- model: Model name used for pricing
"""
# Get pricing for the specified model
pricing = get_model_pricing(model)
total_cached_tokens = 0
total_prompt_tokens = 0
total_output_tokens = 0
total_uncached_tokens = 0
for usage in usage_list:
total_prompt_tokens += usage.prompt_tokens
total_output_tokens += usage.completion_tokens
if hasattr(usage, "prompt_tokens_details") and hasattr(usage.prompt_tokens_details, "cached_tokens"):
cached_tokens = usage.prompt_tokens_details.cached_tokens
if isinstance(cached_tokens, int):
total_cached_tokens += cached_tokens
else:
total_cached_tokens += 0
else:
cached_tokens = 0
total_uncached_tokens += usage.prompt_tokens - cached_tokens
# Calculate costs using model-specific pricing
cost_uncached = (total_uncached_tokens / 1_000_000) * pricing.input_price
cost_cached = (total_cached_tokens / 1_000_000) * pricing.cached_input_price
cost_output = (total_output_tokens / 1_000_000) * pricing.output_price
total_cost = cost_uncached + cost_cached + cost_output
print(f"Estimated Cost: ${total_cost:.4f}")
return {
"total_cached_tokens": total_cached_tokens,
"total_prompt_tokens": total_prompt_tokens,
"total_output_tokens": total_output_tokens,
"total_uncached_tokens": total_uncached_tokens,
"cost_uncached": cost_uncached,
"cost_cached": cost_cached,
"cost_output": cost_output,
"total_cost": total_cost,
"model": model,
}
AI Essay Evaluator