Gemini 2.5 API: How to Use Google's 1 Million Token Context Window

Google's Gemini 2.5 Pro and Gemini 2.5 Flash offer the largest context windows available in production AI models — up to 1 million tokens. That's roughly 750,000 words, or about 10 full-length novels. This unlocks use cases that simply aren't possible with 128K-200K context models.

What Can You Fit in 1M Tokens?

Gemini 2.5 Pro vs Flash

Accessing Gemini 2.5 via OpenAI SDK

You don't need Google's SDK. AIPower wraps Gemini in the standard OpenAI format:

from openai import OpenAI

client = OpenAI(base_url="https://api.aipower.me/v1", api_key="YOUR_KEY")

# Analyze an entire codebase
with open("codebase_dump.txt") as f:
    code = f.read()  # Could be 500K+ tokens

response = client.chat.completions.create(
    model="google/gemini-2.5-pro",  # 1M context
    messages=[
        {"role": "system", "content": "You are a senior code reviewer."},
        {"role": "user", "content": f"Review this codebase for security issues:\n{code}"}
    ],
)
print(response.choices[0].message.content)

Use Case: Codebase Q&A

Load your entire repository into context and ask questions about it. No embeddings, no RAG pipeline, no vector database — just dump the code and ask.

import os

def load_codebase(directory, extensions=(".py", ".ts", ".js")):
    """Load all source files into a single string."""
    files = []
    for root, _, filenames in os.walk(directory):
        for fn in filenames:
            if fn.endswith(extensions):
                path = os.path.join(root, fn)
                with open(path) as f:
                    files.append(f"### {path}\n{f.read()}")
    return "\n\n".join(files)

code = load_codebase("./my-project")
# Now pass 'code' as context to Gemini 2.5 Pro

Use Case: Document Summarization at Scale

Process entire reports, legal contracts, or research papers without chunking:

# Summarize a 200-page annual report
response = client.chat.completions.create(
    model="google/gemini-2.5-flash",  # Flash is cheaper for bulk processing
    messages=[
        {"role": "system", "content": "Summarize this annual report. "
         "Focus on: revenue, growth metrics, risks, and forward guidance."},
        {"role": "user", "content": annual_report_text}  # 150K+ tokens
    ],
)
# Cost: ~$0.02 for input + ~$0.01 for output = ~$0.03 total

When to Use Gemini vs Other Models

• Use Gemini 2.5 Pro when your input exceeds 128K tokens and quality matters.
• Use Gemini 2.5 Flash for high-volume long-context tasks where speed and cost matter more than peak quality.
• Use Claude Opus 4.6 (200K context) for tasks under 200K where reasoning quality is paramount.
• Use Doubao Pro (256K context, $0.06/M) as a budget long-context option.

All these models are available through a single API at aipower.me. Switch between them by changing one parameter. Start with 50 free API calls.