Managing and Reducing Costs › Strategy 4: Batch Similar Requests

from openai import OpenAI

client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

def process_individually(queries):
    """Process each query separately (expensive)."""
    results = []
    total_cost = 0.0
    
    for query in queries:
        response = client.chat.completions.create(
            model="gpt-5",
            max_completion_tokens=100,
            messages=[{"role": "user", "content": query}]
        )
        
        # Estimate cost (rough approximation)
        tokens = response.usage.total_tokens
        cost = (tokens / 1_000_000) * 6.25  # Average of input/output rates
        total_cost += cost
        
        results.append(response.choices[0].message.content)
    
    return results, total_cost

def process_batched(queries):
    """Process all queries in a single call (cheaper)."""
    # Combine queries into a single prompt
    batch_prompt = "Answer each of the following questions concisely:\n\n"
    for i, query in enumerate(queries, 1):
        batch_prompt += f"{i}. {query}\n"
    
    response = client.chat.completions.create(
        model="gpt-5",
        max_completion_tokens=500,
        messages=[{"role": "user", "content": batch_prompt}]
    )
    
    # Estimate cost
    tokens = response.usage.total_tokens
    cost = (tokens / 1_000_000) * 6.25
    
    # Parse the batched response
    result = response.choices[0].message.content
    
    return result, cost

## Test both approaches
queries = [
    "What is Python?",
    "What is JavaScript?",
    "What is Ruby?",
    "What is Go?",
    "What is Rust?"
]

print("Individual processing:")
results_individual, cost_individual = process_individually(queries)
print(f"Cost: ${cost_individual:.4f}")
print()

print("Batched processing:")
result_batched, cost_batched = process_batched(queries)
print(f"Cost: ${cost_batched:.4f}")
print(f"Savings: ${cost_individual - cost_batched:.4f} ({((cost_individual - cost_batched) / cost_individual * 100):.1f}%)")
print()
print("Batched response:")
print(result_batched)