Gemini API Pricing (November 2026)

The Gemini API has a free tier, and using Google AI Studio in the browser costs nothing in every region where it is available. On the free tier, input and output tokens cost nothing on the Flash and Flash-Lite models, up to each model's free rate limits.

Gemini 4 Carbon, Gemini 4 Argon and Gemini 3.1 Pro Preview have no free tier. They run only on the paid tier, at the rates in the right-hand column below.

US dollars per million tokens, paid Standard tier, prompts up to 200,000 tokens. Free-tier status from Google's pricing page.

The free tier has a cost of its own. Google uses free-tier prompts and responses to improve its products, and it does not do so on the paid tier. Keep private or customer data on a paid project.

Free limits apply per project, not per API key, and the daily request quota resets at midnight Pacific time. Google shows each project's current limits in AI Studio rather than in its docs. The glossary entry on free tiers compares the Gemini offer with other providers.

Implicit caching is on by default for Gemini 2.5 and newer models. When a request repeats a prefix Google has already processed, the repeated tokens bill at the cache rate, roughly a tenth of the input rate. The prefix must be at least 4,096 tokens on Gemini 4 Carbon and Argon, 3.8 Flash, 3.6 Flash and 3.1 Pro Preview, and 2,048 on Gemini 2.5 Flash and 2.5 Pro.

Explicit caching lets you store a prompt for a set time and reuse it on purpose. It adds a storage fee per million cached tokens per hour: $0.50 on Gemini 3.8 and 3.6 Flash through December 31, 2026, $1 on the Flash-Lite models and $4.50 on Pro models. Gemini 4 Carbon, Gemini 4 Argon and the Interactions API support implicit caching only.

US dollars per million tokens, input / output, prompts up to 200,000 tokens. The Gemini 4 rates run through November 30, 2026, and the 3.8 and 3.6 Flash rates through December 31, 2026.

• Batch runs requests asynchronously at half the Standard rate. Each model has its own limit on queued Batch tokens, and the limit rises with the usage tier.
• Flex also costs half, in exchange for variable latency and best-effort processing. Google still labels Flex a preview.
• Priority costs 1.8 times Standard and has rate limits of its own, 0.3 times the standard limit by default.

Google also shows a Contributor tier, an invite-only private preview where your prompts are used to improve Google's products. On November 7, 2026 its card read $1 input and $5 output for Gemini 4 Argon, with no end date; Gemini 4 Carbon has no Contributor card. The glossary explains prompt caching, the batch API and service tiers across providers.

Four Gemini price changes land within two months. Each one is published on Google's pricing or deprecation pages with a date.

From Google's pricing page, read November 7, 2026. Batch, Flex and Priority rates change by the same factor.

On January 1, 2027, Gemini 3.8 Flash moves to $1.50 input and $7.50 output per million tokens. Gemini 3.1 Flash-Lite keeps its card until Google shuts it down on May 7, 2027, and Google names Gemini 3.5 Flash-Lite as its replacement.

Budgets built on today's rates need a second line for 2027. A workload that costs $100 a month on Gemini 3.8 Flash in December will cost about $200 in January at the same volume. Moving bulk jobs to Batch or Flex holds the bill near its December level.

Gemini 3.1 Pro Preview charges more once a prompt passes 200,000 tokens. Input goes from $2 to $4, output from $12 to $18, and cache reads from $0.20 to $0.40. Gemini 4 Carbon, Gemini 4 Argon and the Flash and Flash-Lite models charge one rate across their full 1 million token window.

Output prices include thinking tokens, so a model that reasons at length bills those tokens as output. Audio input costs more than text on some models, such as Gemini 3.1 Flash-Lite and Gemini 3 Flash Preview.

Built-in tools add their own charges on the paid tier:

• Grounding with Google Search on Gemini 3 and newer models: 5,000 free requests a month across all those models, then $14 per 1,000, with a default cap of 1,500 requests a day.
• Grounding with Google Maps: the same 5,000 free requests a month, then $14 per 1,000.
• URL context, code execution and computer use: no separate fee; the content and output bill as tokens at the model's rates.
• File search: indexing bills at Gemini Embedding rates, and retrieved text bills as input tokens.

The Gemini API also sells image, video, speech, music and embedding models on the same account. Their prices are in the image generation, video generation, text-to-speech and embedding cost calculators. The long-context pricing entry compares Google's 200K step with other providers.

A Google AI plan and the Gemini API are billed separately. A plan pays for the Gemini app, Gemini in Google apps and the creative tools. The API bills every token through AI Studio, and no Google AI plan includes API credit.

US plan prices from our plan sheet, checked 2026-11-07 on gemini.google.

The paid API tier starts when you link a billing account in AI Studio and prepay at least $5. Unused prepaid credit expires after 12 months and is not refundable.

When the prepaid balance reaches $0, every API key on that billing account stops working. Requests fail with HTTP 402 until you add credit, so top up before a production balance runs out.

A plan suits daily use of the Gemini app or Gemini in Gmail and Docs. The API suits software you build, and light use often costs less than a plan. The subscription vs API calculator runs the numbers for your volume, and Gemini vs ChatGPT compares the consumer plans.

Each project sits in a usage tier, and its rate limits rise with the tier. Tiers 2 and 3 count total spending on Google Cloud services by the linked billing account, not only Gemini API spending.

From Google's rate-limit page, read November 7, 2026.

The spend limit is a rolling 10-minute window that guards against runaway bills. Hitting it returns HTTP 429 RESOURCE_EXHAUSTED, as does exceeding a request or token limit. A project moves up a tier automatically once it meets the criteria.

Google sells the same models on Gemini Enterprise Agent Platform, formerly Vertex AI, billed through Google Cloud. Google says those prices may differ from the Gemini API rates on this page. The glossary explains rate limits across providers.