Estimate your monthly Google Gemini API spend by token usage and request volume.
| Model | $/1M input | $/1M output | Context |
|---|---|---|---|
| Gemini 3.1 Pro (preview) | 2.00 | 12.00 | 1M |
| Gemini 3.6 Flash | 1.50 | 7.50 | 1M |
| Gemini 3.5 Flash-Lite | 0.30 | 2.50 | 1M |
| Gemini 3.1 Flash-Lite | 0.25 | 1.50 | 1M |
Gemini's headline feature is its 1M-token context window — far larger than most rivals — which makes it attractive for long-document and whole-codebase workloads. Flash is the budget tier for high-volume, latency- sensitive tasks; Pro for harder reasoning.
The large context window is Gemini's headline feature and the easiest way to run up a bill. Tokens are charged on every request, so a workflow that habitually sends hundreds of thousands of tokens costs orders of magnitude more per call than one that retrieves the relevant few thousand — for the same question.
Google has also priced long requests differently from short ones at various points, with a higher rate above a threshold. Whether that applies to the model you are using, and where the threshold sits, is worth checking on the official pricing page before designing around a very large context.
Context caching is the counterweight. When the same large document or corpus is queried repeatedly, caching it charges a reduced rate for reuse instead of full price on every call. It usually carries a storage component priced by duration, so it pays off for content queried many times and not for one-off requests.
The same models are reachable through the Gemini API directly and through Vertex AI on Google Cloud. The model capabilities match, but billing, quotas and available regions differ, and rates are not always identical between the two. The figures above track the direct API — if you are billing through a Google Cloud project, confirm against Vertex AI's own pricing page.
The Flash tiers are built for volume: high throughput, low latency, and cheap enough to sit in a request path that runs constantly. Pro tiers cost more per token and earn it on harder reasoning. As with any provider, the useful test is not which model is better in general but whether the cheaper one fails on your actual task — for extraction, classification and summarisation it frequently does not.
Gemini 3.1 Flash-Lite, at $0.25/$1.5 per million tokens. Gemini 3.1 Pro (preview) is the priciest listed here at $2/$12.
Use the Token Counter, then plug the numbers in above.
Not necessarily, and preview pricing can change when a model reaches general availability. Anything marked preview should be treated as provisional for budgeting: the capability may be exactly what you want, but the rate you plan around today is not a commitment.
Google offers free usage with rate limits on some models, which is enough for prototyping and evaluation but not for a product with real traffic. The limits and the terms — including whether inputs may be used to improve the service — differ between the free and paid tiers, so check both before building on the free one.
Compare across providers: LLM API Cost Calculator · Claude · GPT