Gemini 2.5 Pro — how a 1M context actually helps

A million tokens is roughly 750,000 words — the entire Lord of the Rings trilogy plus the Hobbit, with room to spare. Gemini 2.5 Pro holds that in working memory at $1 in and $10 out per million tokens. That is cheap enough to actually use. The question is: what do you do with it?

The trap: you can dump too much

Just because it all fits does not mean you should paste it all. Every token costs money going in and distracts the model on the way out. If the answer is in chapter 3, do not send chapters 1-20.

• Chunk first, summarize, then send just the relevant chunks.
• For true long-context tasks (finding a pattern across the whole thing) send the whole thing but be specific about what you need.
• Use Gemini's grounding (search) for facts the model could not possibly know.

import google.generativeai as genai

genai.configure(api_key=os.environ["GEMINI_API_KEY"])
model = genai.GenerativeModel("gemini-2.5-pro")

with open("full_codebase_dump.txt", "r") as f:
    codebase = f.read()  # ~400k tokens of Python

resp = model.generate_content(
    [
        codebase,
        "Find every place where user input reaches the database without validation. "
        "Give me file:line and the risk severity."
    ],
    generation_config={"temperature": 0.1}
)
print(resp.text)