Gemini API Price Increase: What Doubling Does to Your Monthly Bill

What Google is changing, and when

Google's Gemini API pricing page now shows two prices for three models: the current rate and a higher rate from a fixed date. Each line doubles. Nothing else about the models changes on those dates.

Batch, Flex and Priority prices move by the same factor on the same dates. Gemini 3.8 Flash also doubles its context cache storage fee, from $0.50 to $1.00 per million tokens per hour. The rates come from Google's pricing page, and each change is listed with its source in our model price monitor. Current rates are on the Gemini 4 Argon and Gemini 3.8 Flash pricing pages.

Two workloads, before and after

Because every rate doubles, the bill doubles whatever your mix of input, output and cached tokens. The examples are here to put dollar figures on that. Both are made up, so swap in your own volumes.

• Chat assistant: 200,000 requests a month, about 1,500 input and 400 output tokens each. That is 300 million input and 80 million output tokens.
• Document Q&A: 50,000 questions a month over a 10,000-token handbook that stays in the cache, plus 2,000 fresh input tokens and 300 output tokens per question. That is 500 million cached, 100 million uncached input and 15 million output tokens.

The document workload assumes the handbook is billed at the cached input rate on every call, which is the best case. Gemini's implicit caching is on by default but gives no guarantee of a hit, so a lower hit rate moves more of that input to the full rate. Explicit caching on 3.8 Flash adds the hourly storage fee, which these numbers leave out.

Models that cost the same as Gemini today

If the new rate breaks your budget, the first question is which models cost what Gemini costs now. Here is the chat assistant workload priced at current list rates on models in the same price range. These figures update when those providers change prices.

Staying inside Gemini is the smallest change. Gemini 3.5 Flash-Lite has no scheduled increase on Google's page, and Google will route Gemini 2.5 Flash requests to it from December 1. On our measured benchmark, Flash-Lite got every everyday task right within three attempts but missed 6 of 270 hard items, so check it against your harder cases first.

If you still call Gemini 2.5 Pro

Gemini 2.5 Pro is deprecated. Google says that from December 1, 2026, requests to gemini-2.5-pro route automatically to Gemini 3.8 Flash and bill at Gemini 3.8 Flash rates. On the chat workload that is a drop first and a rise a month later: $1,175 a month on 2.5 Pro's short-context rates, $525 in December on 3.8 Flash, and $1,050 from January 1.

The model behind your requests changes on December 1 even if your code does not. Test your prompts on Gemini 3.8 Flash before then, because answers, formatting and refusals can all shift.

Check your own numbers before the date

• Pull last month's token counts per model from your usage report, split into input, cached input and output.
• Multiply by the new rates above, or enter them in the calculator below. For these three models the result is simply twice today's bill.
• Pick one or two cheaper candidates and run 50 to 100 of your real tasks through each. Count the answers you would ship as they are.
• Compare cost per successful task: model cost per call divided by that success rate. A 10% cheaper model with 85% success instead of 95% is not cheaper.

The model switch cost calculator prices one workload on two models and takes a success rate for each. Save the workload there and it shows the scheduled Gemini price and a 12-month difference at the new rate, using official sources. It is stored in your browser only and nothing alerts you, so open it again after the date. If retries are part of your setup, the retry and fallback cost calculator shows what extra attempts add per successful task.

FAQ