Gemini API pricing at a glance
The lowest headline token rate is not the cost of every request. Model, input length, output length, thinking tokens, speed tier, caching and tools can all change the final bill.
Selected Gemini Developer API rates checked November 7, 2026| Route | Free tier | Standard paid rate | What changes the bill |
|---|
| Gemini 4 Argon input | Not available | $2.00 / 1M tokens through Nov 30, 2026; $4.00 from Dec 1 | Input tokens, including some tool instructions and retrieved context |
|---|
| Gemini 4 Argon output | Not available | $10.00 / 1M through Nov 30; $20.00 from Dec 1 | Visible output plus billed thinking tokens |
|---|
| Gemini 4 Argon context cache | Not available | $0.10 / 1M through Nov 30; $0.20 from Dec 1 | Implicit caching only at this rate; storage may be separate |
|---|
| Gemini 4 Argon Batch | Not available | $1.00 input and $5.00 output / 1M through Nov 30 | Asynchronous batch processing; rates also double on Dec 1 |
|---|
| Gemini 3.8 Flash | Free of charge | $0.75 input and $3.75 output / 1M through Dec 31 | Free availability, rate limits and data-use terms differ |
|---|
| Google Search grounding | Limited by model and quota | 5,000 shared requests/month, then $14 / 1,000 | Each search query can be billed, subject to the daily request cap |
|---|
How Gemini API token billing works
Google bills models by tokens rather than by chat message. Input tokens cover the prompt and context sent to the model; output tokens cover the response and any reasoning tokens the model generates internally.
- ·Long prompts, documents, images and retrieved tool context can increase input tokens.
- ·Thinking tokens are included in the output price shown for reasoning models.
- ·Grounding tools can add per-request charges beyond model input and output.
- ·Context caching can reduce repeated input cost, but cache storage and eligibility have separate terms.
- ·Batch and Flex processing trade latency for lower rates; Priority processing costs more.
What is included in the Gemini API free tier?
The free tier is intended for testing and small projects. It offers limited access to certain models, free input and output tokens, and Google AI Studio access, but it is not a complete preview of every paid model or production feature.
- ·Gemini 4 Argon is explicitly unavailable on the free tier.
- ·Rate limits and supported models can differ from the paid tier.
- ·Free-tier content may be used to improve Google products under the applicable terms.
- ·Paid-tier access adds higher limits, Context Caching, Batch API and advanced models.
- ·An API key being valid does not guarantee that every model has free quota.
Why Gemini 4 Argon costs more from December 1
Google's current pricing page publishes introductory Gemini 4 Argon rates through November 30, 2026 and higher standard rates beginning December 1. Applications that forecast spend from the current rate will underestimate December usage.
Gemini 4 Argon standard price transition| Charge | Through Nov 30, 2026 | From Dec 1, 2026 |
|---|
| Input | $2.00 / 1M tokens | $4.00 / 1M tokens |
|---|
| Output, including thinking | $10.00 / 1M tokens | $20.00 / 1M tokens |
|---|
| Context caching | $0.10 / 1M tokens | $0.20 / 1M tokens |
|---|
| Batch input | $1.00 / 1M tokens | $2.00 / 1M tokens |
|---|
| Batch output | $5.00 / 1M tokens | $10.00 / 1M tokens |
|---|
| Priority input | $3.60 / 1M tokens | $7.20 / 1M tokens |
|---|
| Priority output | $18.00 / 1M tokens | $36.00 / 1M tokens |
|---|
A worked Gemini API cost example
A request batch using 2 million input tokens and 500,000 output tokens costs $9.00 at the current Gemini 4 Argon standard rates. From December 1, the same token volumes cost $18.00 before tools, cache storage or taxes.
Gemini 4 Argon cost for 2M input + 0.5M output tokens| Processing route | Through Nov 30 | From Dec 1 |
|---|
| Input charge | 2 × $2.00 = $4.00 | 2 × $4.00 = $8.00 |
|---|
| Output charge | 0.5 × $10.00 = $5.00 | 0.5 × $20.00 = $10.00 |
|---|
| Estimated total | $9.00 | $18.00 |
|---|
| Batch equivalent | $4.50 | $9.00 |
|---|
Free tier vs paid tier: what changes
The paid tier is the production path: it raises rate limits, unlocks advanced models and adds tools that are unavailable or restricted on the free tier. Exact quotas depend on the model and current Gemini API terms.
- ·Paid-tier content is not used to improve Google products, subject to the API terms.
- ·Context Caching and Batch API are paid-tier features.
- ·Advanced models and higher rate limits are intended for production applications.
- ·Billing is attached to a Google Cloud billing account and project.
- ·Budget alerts are safeguards, not hard spending caps unless a separate control enforces them.
Does Google AI Pro or Ultra include Gemini API credits?
No. Google AI Plus, AI Pro and AI Ultra are consumer plan products. The Gemini Developer API uses separate projects, API keys, quotas and billing; a consumer subscription does not normally convert into reusable API tokens.
Consumer subscription versus Gemini Developer API| Purchase | Primary use | Account and billing | What it does not provide |
|---|
| Google AI Plus, Pro or Ultra | Consumer Gemini app and Google AI features | Google account subscription in local currency | A general-purpose API credit balance |
|---|
| Gemini Developer API | Application requests and production workloads | AI Studio or Google Cloud project with API billing | The consumer storage and app bundle |
|---|
| One-month Google AI top-up | A time-limited consumer membership on one account | Independent one-time purchase through Gemini Pricing | API keys, Cloud credits or project quota |
|---|
| Gemini Enterprise | Organization workspace, agents and governance | Business or enterprise seat agreement | Consumer API-metered pricing unless separately purchased |
|---|
How to estimate and reduce API spend
Start with measured token counts rather than message counts. Separate development and production projects, cap prompts and outputs, monitor each model and test lower-cost routes before moving a workload to priority processing.
- ·Measure prompt, cache, output and thinking tokens for representative requests.
- ·Use Batch or Flex when the workflow can tolerate higher latency.
- ·Cache stable instructions only when the model and workload qualify.
- ·Avoid repeatedly sending the same large documents or system instructions.
- ·Set project budgets, quota alerts and application-level usage limits.
- ·Keep API keys server-side and rotate any key exposed in logs or client code.
Which Gemini API pricing route should you choose?
Use the free tier to evaluate supported models and small prototypes, Standard processing for interactive production requests, Batch or Flex for delay-tolerant workloads, and Priority only when latency or reliability justifies the premium.
- ·Choose Flash models when the workload fits a lower-cost model.
- ·Choose Gemini 4 Argon when its reasoning and long-decode capabilities are required.
- ·Use Gemini Enterprise or a Cloud agreement for organizational governance and support.
- ·Use a consumer top-up only for interactive Gemini use, never to fund API requests.