Gemini API prices depend on the model and token mix
Gemini API charges are separate from a Gemini app subscription. Google’s documentation lists these prices per 1 million tokens:
| Model | Model ID and status | Input | Output | Price support |
|---|---|---|---|---|
| Gemini 4 Argon | gemini-4-argon — Stable |
$2 through November 30, 2026; $4 from December 1, 2026 | $10 through November 30, 2026; $20 from December 1, 2026 | Google documents the rates and dates. |
| Gemini 3.8 Flash | gemini-3.8-flash — Stable |
$0.75 through December 31, 2026; $1.50 from January 1, 2027 | $3.75 through December 31, 2026; $7.50 from January 1, 2027 | Google documents the introductory and standard rates. |
| Gemini 3.5 Flash-Lite | gemini-3.5-flash-lite — Stable |
$0.30 reported | $2.50 reported | These rates are reported by God of Prompt; I could not confirm them in the retrieved Google pricing material. Treat them as unverified and check Google’s Gemini API pricing page before budgeting. |
Google’s documentation describes the Argon rates as introductory through November 30, and Flash’s as introductory through December 31. The listed rates use dollar signs per million tokens; the retrieved excerpts do not spell out the currency. Prices checked November 8, 2026. Google describes paid API access as intended for production applications. (Argon pricing; Flash pricing; Google API pricing)
Estimate a production budget
For a text-token estimate, calculate input and output separately:
Estimated charge = (monthly billable input tokens ÷ 1,000,000 × input rate) + (monthly billable output tokens ÷ 1,000,000 × output rate).
To estimate monthly totals, multiply requests per month by average billable input and output tokens per request. For example, 10,000 requests averaging 100 input tokens and 10 output tokens each use 1 million input tokens and 100,000 output tokens.
At Google’s documented Argon rates, that workload’s estimated token charge is $3.00 through November 30, 2026 ($2 input + $1 output), or $6.00 from December 1 ($4 input + $2 output). This is an estimate, not a guaranteed invoice. For models that use thinking, include billable thought tokens in the output estimate; Google says pricing is based on the full thought tokens generated, even when the API returns only a summary. (Google’s thinking documentation)
Google lists Batch API as offering a 50% cost reduction on its paid tier. Apply that only when your requests qualify under current terms. The example does not estimate caching, tools, or non-text inputs. (Google API pricing)
Which model should you evaluate?
Start with the workload, then test candidate models on representative requests and compare output quality, latency, and billable token use. Google describes Argon as its flagship reasoning model for complex software engineering, mathematical and scientific problem-solving, and autonomous agents. Google describes Flash as designed for long-horizon software engineering, autonomous agents, and complex enterprise workflows. Flash-Lite is described as a low-latency, cost-effective model for high-throughput execution, subagent tasks, and document parsing. These descriptions can help build a shortlist, but your own evaluation is needed to establish fit. (Argon model page; Flash model page; Flash-Lite model page)
For each candidate, calculate expected monthly input and output separately using a rate confirmed for the exact model and API channel you plan to use. Before launch, recheck current prices and model availability in Google’s documentation; do not treat an unconfirmed third-party rate as an official rate.