Google introduced Gemini 4 Hydrogen on November 12 at $1.50 per million input tokens and $7.50 per million output tokens. At those rates, a batch consuming 100,000 input tokens and 10,000 output tokens costs $0.225 for those tokens. The expensive mistake is treating a total token count as if every token had the same price.
Work through a support-ticket example
Suppose you summarize 1,000 support tickets. Each request uses 2,000 input tokens and produces 200 billable output tokens. That is 2 million input tokens ($3) and 200,000 output tokens ($1.50), for a token bill of $4.50. These are illustrative usage assumptions, not measurements of a customer deployment.
Now keep the source tickets unchanged but ask for 600 billable output tokens per response. Output rises to 600,000 tokens ($4.50). Total token cost becomes $7.50. The extra $3 bought longer responses; it did not buy any additional source material.
Measure usefulness alongside cost
Run a representative sample before projecting a month of traffic. Include short tickets, long threads and at least one ticket where the customer's last message reverses an earlier request. Check whether the summary preserves that change. Log actual billable usage, retries and how often a reviewer must rewrite the answer.
Google says its listed rates are flat across the model’s million-token context window. A large window still lets you send a larger billable input. Use the number of turns your application really sends, including repeated context. Budget tool charges, taxes or other services separately where applicable.
Sources: Google’s Hydrogen launch and published rates. Image: Google.
