Gemini API Pricing (November 2026)

Google bills the Gemini Developer API (Google AI Studio and API keys) by the million tokens in US dollars, with separate Free, Paid, and Enterprise product cards—not one flat rate for every request, and not the same wall as consumer Gemini App or Google One seats. On Paid Standard, current headline rows include gemini-4-argon at $2 input / $10 output per 1M tokens through November 30, 2026, then $4 / $20 from December 1, 2026; gemini-3.8-flash and gemini-3.6-flash at $0.75 / $3.75 through December 31, 2026, then $1.50 / $7.50 from January 1, 2027; and lighter Flash rows from about $0.25–$0.30 input. Output prices include thinking tokens. Figures below follow Google’s Gemini API pricing and deprecations documentation as of this write-up (2026-11-03). For reference only. Consumer Gemini App / Google One subscription ladders are a different product surface—this page stays on Developer API token rates.

Free, Paid, and Enterprise

The pricing page opens with three product cards.

Free is oriented to getting started and smaller projects. Some models list Free of charge input and output tokens inside rate limits; Google AI Studio access is included; and the card states that content is used to improve Google’s products (see the official Terms footnote). Free is not unlimited: exact RPM / TPM / daily quotas vary by model and are shown in AI Studio and the rate-limits docs rather than as a single public number for every ID.

Paid is oriented to production and higher volume. The card lists higher rate limits, access to context caching, Batch API (50% cost reduction), access to Google’s most advanced models, and content not used to improve products. Upgrading is a billing step on the Developer API path—not a ChatGPT-style personal seat.

Enterprise points to deployments on Gemini Enterprise Agent Platform. The card stacks Paid capabilities plus optional dedicated support, advanced security and compliance, provisioned throughput, volume-based discounts, and related ML-ops surfaces. Pricing is Contact Sales—no seat dollars are listed here for that cloud path. Notes on the Developer API pricing page also say Enterprise Agent Platform prices may differ from the list here.

Argon Priority and Contributor tabs exist on some model sections. Priority is a non-default, higher-priced optional lane (Argon example: about $3.60 / $18 through November 30, then $7.20 / $36). Contributor is described as a private preview / invite path (Argon example around $1 / $5, with “used to improve” marked Yes on that tab). Neither is the default Standard map used for comparisons here.

Standard Paid rates (core chat models)

Prices below are Paid Standard, USD per 1M tokens, unless noted. Free-of-charge cells mean the Free tier lists free tokens for that model within limits—not that production traffic is uncapped.

gemini-4-argon is the flagship reasoning row and is not on Free. Google’s callout says every token the model processes is counted and billed, including thinking tokens and internal tool system instructions and function declarations when using grounding. Context caching on Argon Standard is implicit only.

gemini-3.8-flash and gemini-3.6-flash share the same Standard promo ladder on the live page: $0.75 / $3.75 through year-end 2026, then $1.50 / $7.50. Cache storage on those Flash rows steps from $0.50 to $1.00 per million cached tokens per hour on the same date. Both list Free of charge on Free for the main input/output cells.

gemini-3.5-flash-lite is the common low-cost multimodal Flash-Lite on Paid at $0.30 / $2.50. gemini-3.1-flash-lite sits slightly lower on text/image/video input ($0.25) with a separate $0.50 audio input cell and $1.50 output.

gemini-3.1-pro-preview uses a 200k prompt-size split rather than Argon’s flat promo pair: $2 / $12 at or under 200k input tokens, $4 / $18 above. The customtools sibling is priced on the same orientation on the pricing page.

Deprecated 2.5 rows remain on the table for existing API users until the December 1, 2026 routing cutover described next. Matching model ID, Free vs Paid, and the Standard tab—not a neighboring Priority cell—matters before estimating a bill.

Batch (and Flex) in one pass

The Paid product card advertises Batch API (50% cost reduction). On many model tabs, Batch (and where present, Flex) cells land at about half of Standard. Examples from the live tabs: Argon Batch / Flex input $1→$2 and output $5→$10 on the same November 30 / December 1 cut; 3.8 / 3.6 Flash Batch input $0.375→$0.75 and output $1.875→$3.75 on the December 31 / January 1, 2027 cut. The Standard wall above comes first; Batch/Flex is about half Standard on most listed models—verify the Batch tab for the ID—without a third full price wall.

Search Grounding and other tools

Tool prices sit on top of model token rates. For Gemini 3 and newer Search Grounding on Paid, Google lists about 5,000 free search requests per month shared across Gemini 3 and newer models, then about $14 per 1,000 requests. On the Developer API there is also a default daily cap of about 1,500 RPD per project, resetting at midnight Pacific Time. One customer request may trigger multiple Search queries; billing is per search query performed. Free-tier cards mark Gemini 3+ Search as Not available on several rows (AI Studio may still allow limited testing—API Free is not a production Grounding grant).

Gemini 2.5 Search uses a different Paid shape: about 1,500 RPD free (Flash / Flash-Lite shared on several rows), then about $35 per 1,000 grounded prompts. The 3+ and 2.5 columns stay on separate lines in cost models.

Code execution, URL context, and computer use are generally billed as ordinary tokens at the selected model’s rates (session runtime for code execution is not a separate clock charge on the tools table). File search embeddings are exemplified at Gemini Embedding $0.15 / 1M (or Embedding 2 rates), with retrieved document tokens billed at the model rate. Deep Research and managed agents charge underlying model tokens plus tool fees—no separate “agent seat” dollar is listed here.

Deprecated 2.5 routes (December 1, 2026)

Google’s deprecations page and the pricing caution blocks agree on the cutover:

• gemini-2.5-pro: deprecated; no longer selectable in AI Studio for new choice. Effective December 1, 2026, API requests automatically route to gemini-3.8-flash and bill at 3.8 Flash rates.
• gemini-2.5-flash: deprecated; effective December 1, 2026, API requests automatically route to gemini-3.5-flash-lite and bill at 3.5 Flash-Lite rates.

Until that date, the 2.5 Standard cells above still describe how remaining eligible traffic is listed. After routing, the destination model’s Standard (or Batch) tab is what the invoice follows—not the legacy 2.5 cell.

Image, Live, TTS, and other modality tables

The same pricing URL carries separate tables for image (Nano Banana family), Live / native audio, TTS, transcription, embeddings, robotics, and related rows—some priced per image, per minute, or per track rather than only per 1M text tokens. Those full walls are not reprinted here. Notes worth carrying forward: DOCUMENT (for example PDF) modality tokens bill at the image token rate; agentic video understanding may use far fewer input tokens than raw duration would suggest (the Notes text mentions reductions on the order of up to about 88% for long-form video depending on query and sampling—that figure is an official orientation, not a guaranteed savings factor).

Rate limits, Studio, and billing hygiene

Free-of-charge model cells are still gated by rate limits that differ by model and project. Google’s rate-limits documentation points teams to the live dashboard rather than a single frozen RPM table in prose. Google AI Studio usage is described as free of charge in available regions, which is separate from upgrading a Developer API project to Paid. Hitting a limit typically means waiting for reset, moving to Paid, or changing model—not an implied human grant of unlimited quota.

For actual invoices or how quickly a given project burns quota, we haven't logged in and reconciled a live bill line by line. Feedback via the site footer contact is welcome from readers who have.

Same-tier API neighbors

On OpenAI’s current API list, OpenAI API Pricing (November 2026) puts flagship Standard short-context rows such as gpt-6.1-sol near $2 / $10 per million input/output—close to Gemini 4 Argon’s promo $2 / $10 headline, though tokenizers, cache rules, Batch tabs, and thinking-token inclusion differ. Anthropic’s Claude API Pricing (October 2026) remains the parallel Claude Developer wall. Equal dollars per million tokens are not equal dollars per task; compare on a fixed prompt after counting tokens on each vendor’s model ID.

App-side Argon usage windows and Gemini CLI daily request caps sit on other meters—see Gemini 4 Argon Usage Limits in the Gemini App, Antigravity and API (October 2026) and How Gemini CLI Usage Limits Work for Each Plan (October 2026).

Separating this map from nearby products

This guide lists Gemini Developer API token and tool prices from Google’s pricing and deprecations pages: Free / Paid / Enterprise cards, the core Standard chat wall, Batch as roughly half Standard, Gemini 3+ vs 2.5 Search Grounding, and the 2026-12-01 2.5→3.x auto-routes. Vertex AI cloud quotes, Enterprise seat sheets, and consumer Gemini App subscription ladders sit on other surfaces; this map stays on Developer API token and tool rates. Tables change—Google’s live pricing docs govern when a cell disagrees with any static copy. For reference only.