Google's current Gemini API model lineup
Google's Gemini API currently prices four models spanning a wide range. Gemini 3.1 Pro sits at the top as the flagship, priced well above the rest of the lineup. Unusually among the providers covered here, its rate isn't flat: prompts over 200,000 tokens are billed at a higher rate, $4 input / $18 output per million tokens, versus the lower rate below that threshold. Gemini 3.5 Flash is the mid-tier option, priced below Pro but still well above the two Flash-Lite models. Gemini 3.1 Flash-Lite and the older Gemini 2.5 Flash-Lite both remain in current pricing at the budget end of the lineup, with Gemini 2.5 Flash-Lite priced lower still, a reminder that Google keeps older, cheaper generations available alongside newer ones rather than retiring them immediately.
How Gemini's pricing structure works
Every Gemini model bills input and output tokens at separate rates, with output priced higher throughout the lineup, consistent with the rest of the market. What sets Gemini 3.1 Pro apart is its context-length-dependent pricing: below 200,000 tokens of prompt, you pay the standard rate; above that, both input and output shift to the higher tier. This is worth building into your calculation deliberately rather than assuming a single flat rate, especially if your workload sits right around that threshold, a prompt that occasionally crosses 200K tokens will occasionally bill at nearly double the rate, which can be an unpleasant surprise if your budgeting only accounted for the lower number.
The two-tier Flash-Lite lineup
Google currently prices both Gemini 3.1 Flash-Lite and Gemini 2.5 Flash-Lite, with the older 2.5 generation priced lower on both input and output tokens. This is a genuinely useful option if your workload is cost-sensitive and doesn't need whatever improvements the 3.1 generation brings; it's worth testing both against your actual traffic rather than assuming the newer model is the right default purely because it's newer. For workloads where quality differences between the two are negligible for your use case, the older, cheaper Flash-Lite can meaningfully lower a high-volume bill.
Cutting your Gemini API bill
The clearest lever is tier selection: route simple, high-volume requests to a Flash-Lite model rather than Gemini 3.1 Pro, and test whether Gemini 2.5 Flash-Lite's lower rate is acceptable for your quality bar before defaulting to the newer 3.1 Flash-Lite. If you're using Gemini 3.1 Pro, watch your prompt length against the 200,000-token threshold specifically, trimming a prompt from just over that line to just under it avoids the higher rate entirely, which is a larger and more mechanical saving than most other cost-cutting moves. Beyond that, the usual levers apply: cap output length, keep system prompts lean, and reuse repeated context where your integration allows it.
A worked example on Gemini 3.1 Pro
Take a 1,000-token prompt with a 500-token answer on Gemini 3.1 Pro at its standard rate ($2 input / $12 output per million tokens, below the 200K-token threshold): input costs 1,000 ÷ 1,000,000 × $2 = $0.002, output costs 500 ÷ 1,000,000 × $12 = $0.006, for $0.008 per call. At 100,000 calls a month that's $800. Push the same prompt past the 200,000-token threshold and the rate jumps to $4 input / $18 output, worth checking directly against your real prompt sizes rather than assuming the lower number holds throughout.
Who the Gemini API suits
Gemini's spread from a very cheap Flash-Lite tier up to a premium, context-aware Pro model suits teams running a genuine mix of traffic, cheap, high-volume classification or extraction on Flash-Lite, and reserving Pro specifically for requests that need either its extra capability or its ability to handle very large prompts. If you want to weigh Gemini against another vendor directly, the all-provider hub calculator puts every provider's models in one comparison view.
Compare Gemini against OpenAI's GPT API pricing, Anthropic's Claude API pricing, or the budget-oriented Mistral API pricing calculator.
