Skip to content

Gemini API Pricing and Quotas: $0.25–$4/1M Input Tokens, Tier Caps

Each Gemini text model's price per 1M tokens, which are free, the $250 / $2,000 / $20,000+ caps on Tiers 1–3, and monthly costs before and after Dec 1 and Jan 1.

•••30 min read•API Guides
Gemini API pricing and quotas: Gemini 4 Argon and Carbon input goes from $2 to $4 per 1M tokens on Dec 1, 2026, Gemini 3.8 Flash from $0.75 to $1.50 on Jan 1, 2027, Gemini 3.1 Flash-Lite costs $0.25, and the monthly tier caps are $250, $2,000 and $20,000+

As of November 14, 2026, Google's Gemini API pricing runs from $0.25 per million input tokens for Gemini 3.1 Flash-Lite to $2.00 for Gemini 4 Argon, Gemini 4 Carbon and Gemini 3.1 Pro Preview, with output from $1.50 to $12.00 per million tokens for prompts up to 200,000 tokens (official pricing page). The cheapest row on the page, Gemini 2.5 Flash-Lite at $0.10, is only open to projects that already used the 2.5 models. Six text models have a free tier, including Gemini 4 Hydrogen and Gemini 3.8 Flash; Carbon, Argon and 3.1 Pro Preview do not. Quotas are counted per project, not per API key: requests per minute, input tokens per minute and requests per day, plus a monthly spending cap of $250 on Tier 1, $2,000 on Tier 2 and $20,000–$100,000+ on Tier 3. Google no longer publishes a fixed per-model RPM table; your project's current limits are on the rate-limit page in AI Studio.

Two dated price changes are already announced. On December 1, 2026, Gemini 4 Argon and Carbon go from $2 / $10 to $4 / $20 per million tokens, and calls to the deprecated gemini-2.5-pro and gemini-2.5-flash start billing at Gemini 3.8 Flash and Gemini 3.5 Flash-Lite rates. On January 1, 2027, Gemini 3.8 Flash and 3.6 Flash go from $0.75 / $3.75 to $1.50 / $7.50.

Gemini API price and quota at a glance (Standard rates per 1M tokens, USD, as of November 14, 2026)

Model (API ID)InputOutput (incl. thinking)Cached inputBatch or Flex (in / out)Free tierScheduled change
Gemini 4 Carbon (gemini-4-carbon)$2.00$10.00$0.10$1.00 / $5.00No$4.00 / $20.00, cache $0.20 from Dec 1, 2026
Gemini 4 Argon (gemini-4-argon)$2.00$10.00$0.10$1.00 / $5.00No$4.00 / $20.00, cache $0.20 from Dec 1, 2026
Gemini 4 Hydrogen (gemini-4-hydrogen)$1.50$7.50$0.075$0.75 / $3.75YesNone announced
Gemini 3.8 Flash (gemini-3.8-flash)$0.75$3.75$0.075 + storage$0.375 / $1.875Yes$1.50 / $7.50 from Jan 1, 2027
Gemini 3.6 Flash (gemini-3.6-flash)$0.75$3.75$0.075 + storage$0.375 / $1.875Yes$1.50 / $7.50 from Jan 1, 2027
Gemini 3.5 Flash-Lite (gemini-3.5-flash-lite)$0.30$2.50$0.03 + storage$0.15 / $1.25YesNone announced
Gemini 3.1 Flash-Lite (gemini-3.1-flash-lite)$0.25 (audio $0.50)$1.50$0.025 + storage$0.125 / $0.75YesEarliest shutdown May 7, 2027
Gemini 3.1 Pro Preview (gemini-3.1-pro-preview)$2.00 (≤200K) / $4.00 (>200K)$12.00 / $18.00$0.20 / $0.40 + storage$1.00 / $6.00 (≤200K)NoPreview, no date
Gemini 3 Flash Preview (gemini-3-flash-preview)$0.50$3.00$0.05 + storage$0.25 / $1.50YesLegacy preview; replacement is 3.6 Flash
Usage tierHow a project gets thereMonthly spend capSpend limit per rolling 10 minutes
FreeActive Google Cloud project (some need verification)NoneNone
Tier 1Link a billing account (new accounts prepay at least $5)$250$10
Tier 2$100 paid and 3 days since the first successful payment$2,000$200
Tier 3$1,000 paid and 30 days since the first successful payment$20,000–$100,000+$200

Tier rules come from Google's billing guide and rate-limit page, as of November 14, 2026. The monthly cap applies to the whole billing account; when it is reached, every linked project pauses until the next month or an approved increase. If your bill goes over a tier's cap, the tier, not the model, is the limit you hit first; the monthly cost tables below show which workloads cross which cap.

For whether your project can use the free tier at all and how much it allows, see Gemini API Free Tier in 2026: What Is Free and What to Check. To read the active RPM, TPM and RPD for each model and diagnose a 429, see Gemini API rate limits by tier.

Understanding Gemini API Quotas

A Gemini API quota is a limit on one project, and a request fails when it exceeds any single one of them. Your budget is a separate limit: a project can stay inside every rate limit and still stop because the billing account reached its monthly cap or ran out of prepaid credit. Google's rate-limit documentation defines the dimensions below, as of November 14, 2026.

Requests Per Minute (RPM) is the number of API calls a project may make in a minute. If your RPM limit is 20, the 21st request inside a minute fails even when every other limit has room. The error is HTTP 429 RESOURCE_EXHAUSTED; retry with exponential backoff rather than immediately.

Tokens Per Minute (TPM) counts input tokens per minute. Long prompts use it up first: a single 100,000-token document uses 100,000 of the minute's input allowance, however short the answer is. Output tokens do not count toward this limit, but they still cost money and still count toward the spend limits described below.

Requests Per Day (RPD) is the daily ceiling, and it resets at midnight Pacific Time. RPD is the limit most free-tier projects hit first. The sharp cut to free-tier daily limits in December 2025 is described further down.

Images Per Minute (IPM) applies only to models that generate images (the Nano Banana family). Some models also have a tokens-per-day (TPD) limit, and built-in tools have their own: Grounding with Google Search on Gemini 3 and newer models has a default cap of 1,500 requests per day per project on the Gemini Developer API.

Spend-based limits are newer. On top of RPM and TPM, Google limits how much a paid project can spend in any rolling 10-minute window: $10 on Tier 1 and $200 on Tiers 2 and 3. Going over also returns 429 RESOURCE_EXHAUSTED. At today's prices, $10 buys 1 million Gemini 4 Argon output tokens, or 2.67 million on Gemini 3.8 Flash. A Tier 1 script that generates long answers with Argon in a loop can therefore hit the spend limit inside ten minutes even though its request rate is low. After December 1, the same $10 buys only 500,000 Argon output tokens.

Three rules decide what number applies to you:

  • Limits belong to the project, not the key. Creating more API keys in the same project does not add capacity; every key draws from the project's limits.
  • The numbers depend on model and tier. Google shows each project's active limits on the rate-limit page in AI Studio and says the specified limits "are not guaranteed." Experimental and preview models get tighter limits.
  • Priority traffic has its own limits. Priority inference is limited by default to 0.3 times the standard rate limit for the same model and tier, and it still counts toward your overall interactive traffic.

The Batch API has separate limits: 100 concurrent batch jobs, a 2 GB input file, 20 GB of file storage, and a per-model ceiling on tokens waiting in the queue (5 million for Gemini 4 Argon and Carbon, 3 million for Hydrogen and 3.8 Flash, 10 million for the Flash-Lite models).

Which Gemini API limit stopped the request: RPM, input TPM and RPD are per project and return 429 RESOURCE_EXHAUSTED; spend limits are $10, $200 and $200 per rolling 10 minutes and $250, $2,000 and $20,000+ per month on Tiers 1 to 3; a prepaid balance of $0 returns HTTP 402

Gemini API Pricing per 1M Tokens by Model (November 2026)

Gemini API prices are per million tokens, and output prices include thinking tokens. For Gemini 4, Google states that every token the model processes is billed, including thinking tokens and the internal tool instructions it adds when you use grounding. Gemini 3.1 Pro Preview and the 2.5 Pro model charge more once a prompt passes 200,000 tokens; the Gemini 4 and Flash models list one rate for the whole context window. All prices below are Standard rates from Google's pricing page as of November 14, 2026.

Gemini 4 Carbon (gemini-4-carbon, generally available since November 10, 2026) is Google's most capable reasoning model. It costs $2.00 input and $10.00 output per million tokens through November 30, 2026, then $4.00 and $20.00 from December 1. Cached input is $0.10, then $0.20. It has no free tier, and caching is implicit only (you cannot create explicit caches).

Gemini 4 Argon (gemini-4-argon, generally available since October 13, 2026) is the faster Gemini 4 model with the same price schedule as Carbon: $2 / $10 until November 30, $4 / $20 from December 1. An invite-only Contributor tier lists Argon at $1.00 / $5.00, but requests on that tier are used to improve Google's products and Google invites eligible users by email.

Gemini 4 Hydrogen (gemini-4-hydrogen, generally available since November 12, 2026) is the Gemini 4 model with a free tier. On the paid tier it costs $1.50 input and $7.50 output per million tokens, $0.75 / $3.75 in Batch or Flex. It adds a fourth xhigh thinking level. All three Gemini 4 models share one 1,048,576-token context window across input, thinking and output.

Gemini 3.8 Flash and 3.6 Flash (gemini-3.8-flash, gemini-3.6-flash) both cost $0.75 input and $3.75 output per million tokens on an introductory price that ends on December 31, 2026; from January 1, 2027 they cost $1.50 and $7.50. 3.8 Flash is the current Flash model; 3.6 Flash is the previous generation. Both have a free tier. Calls to the older IDs gemini-3.7-flash and gemini-3.5-flash are routed automatically to 3.8 Flash and 3.6 Flash and billed at their rates.

Gemini 3.5 Flash-Lite and 3.1 Flash-Lite are the low-cost models. 3.5 Flash-Lite costs $0.30 input (text, image, video or audio) and $2.50 output; 3.1 Flash-Lite costs $0.25 input ($0.50 for audio) and $1.50 output. For a new project, 3.1 Flash-Lite is the cheapest general text model on the price list. Both have a free tier. Google lists May 7, 2027 as the earliest shutdown date for 3.1 Flash-Lite and names 3.5 Flash-Lite as its replacement.

Gemini 3.1 Pro Preview (gemini-3.1-pro-preview) costs $2.00 input and $12.00 output for prompts up to 200,000 tokens, and $4.00 and $18.00 above that. Cached input is $0.20 (or $0.40) plus $4.50 per million tokens per hour of storage. It has no free tier. The old gemini-3-pro-preview ID now points to this model.

Gemini 3 Pro Preview, the top model at the start of 2026 at $2.00 / $12.00, was shut down on March 9, 2026. Requests to gemini-3-pro-preview now go to Gemini 3.1 Pro Preview at the same $2 / $12 list price.

Gemini 2.5 Pro is deprecated and can no longer be selected in AI Studio; API access is limited to projects that used it before. Until November 30 it bills at its old rates:

Context LengthInput (per 1M)Output (per 1M)
≤200K tokens$1.25$10.00
>200K tokens$2.50$15.00

From December 1, 2026, every request to gemini-2.5-pro is routed to Gemini 3.8 Flash and billed at 3.8 Flash rates ($0.75 / $3.75 in December, $1.50 / $7.50 from January 1).

Gemini 2.5 Flash is also deprecated, with the same access limit. Until November 30 it costs:

Input TypeInput (per 1M)Output (per 1M)
Text$0.30$2.50
Image$0.30$2.50
Video$0.30$2.50
Audio$1.00$2.50

From December 1, gemini-2.5-flash requests are routed to Gemini 3.5 Flash-Lite and billed at $0.30 / $2.50, so a text workload's bill does not change; audio input gets cheaper ($0.30 instead of $1.00).

Gemini 2.5 Flash-Lite still costs $0.10 per million input tokens and $0.40 per million output tokens and is not deprecated, but only projects that had already used the 2.5 models can call it (a restriction Google announced on September 18, 2026). A new project gets an error instead of the cheaper rate.

Gemini 2.0 Flash and 2.0 Flash-Lite were shut down on June 1, 2026. Requests to them fail; Google names Gemini 3.6 Flash as the replacement.

What old Gemini model names bill at now

Model ID in your codeStatus (as of November 14, 2026)What you pay
gemini-3-pro-previewShut down March 9, 2026; ID points to 3.1 Pro Preview3.1 Pro Preview: $2 / $12
gemini-2.5-proDeprecated; past users only$1.25 / $10 until Nov 30; 3.8 Flash rates from Dec 1
gemini-2.5-flashDeprecated; past users only$0.30 / $2.50 until Nov 30; 3.5 Flash-Lite rates ($0.30 / $2.50) from Dec 1
gemini-2.5-flash-liteNot deprecated; past users only$0.10 / $0.40
gemini-3.5-flashRouted to 3.6 Flash since Oct 8, 2026$0.75 / $3.75 until Dec 31, then $1.50 / $7.50
gemini-3.7-flashRouted to 3.8 Flash since Oct 8, 2026$0.75 / $3.75 until Dec 31, then $1.50 / $7.50
gemini-2.0-flash, gemini-2.0-flash-liteShut down June 1, 2026Requests fail
gemini-3.1-flash-lite-previewShut down May 25, 2026Requests fail; use gemini-3.1-flash-lite

Status comes from Google's deprecations page and release notes.

Standard, Batch, Flex and Priority prices

The same model costs a different amount depending on how you send the request (optimization guide):

Inference optionPrice vs StandardTypical latencyWhen it fits
StandardFull priceSeconds to minutesNormal interactive calls
Flex50% offMinutes (1–15 minute target), best effortBackground agents, evaluations, data enrichment
Batch API50% offUp to 24 hoursLarge offline jobs
Priority75–100% more (1.8× on Argon, Carbon, Hydrogen and 3.8 Flash)Seconds, not shed under loadLatency-critical production paths
Context cachingCached input billed at 10% of the input rate (a 90% discount) plus hourly storageFaster first tokenLong system prompts or documents reused across calls

Implicit caching is on by default for Gemini 2.5 and newer models, and the discount is applied automatically when a request hits the cache. A prompt must be at least 4,096 tokens long (2,048 on 2.5 models) to be cached. Storage is billed per million tokens per hour: $0.50 on 3.8 Flash until December 31 ($1.00 from January 1), $1.00 on the Flash-Lite models, and $4.50 on 3.1 Pro Preview. The Gemini 4 models support implicit caching only.

Image models and tool charges

Image generation models (Nano Banana) are priced per output image rather than per text token; current rates are in the Gemini image generation cost guide.

Grounding with Google Search is billed by model generation. On Gemini 3 and newer models, the first 5,000 search requests each month are free (shared across all Gemini 3 and newer models), then each 1,000 requests cost $14, with a default limit of 1,500 grounded requests per day per project. On the 2.5 models, the first 1,500 grounded prompts per day are free (shared by Flash and Flash-Lite), then $35 per 1,000. The "1,500" in both rules is a count of requests, not a dollar amount. On the free tier, search grounding is available only on 2.5 Flash and 2.5 Flash-Lite (500 requests per day) and not on Pro models.

Other tools add token charges rather than fees: code execution, URL context and computer use are billed as normal input and output tokens of the model you call; File Search bills the embeddings ($0.15 per million tokens on Gemini Embedding) and the retrieved text as input tokens. Grounding with Google Maps on Gemini 3 and newer models also includes 5,000 free requests a month, then $14 per 1,000.

December 2025 Quota Changes Explained

This section is a record of what happened in December 2025. The free-tier limits in force today are the ones AI Studio shows for your project, and Google's pricing page now lists which models have a free tier at all.

On December 6, 2025, a developer reported on Google's AI developer forum that the free-tier limit for Gemini Flash had dropped from 250 requests per day to 20, overnight and without an announcement. The next day Logan Kilpatrick of Google replied in the same thread that Google had turned down the free tier for some models to free up compute for Gemini 3 Pro and Nano Banana Pro, and described the free tier as best effort and "not likely to persist." Other developers in the thread reported 10 to 20 text requests per day, no free image generation models, and 429 errors in apps that had worked the week before.

What changed in December 2025BeforeAfterSource
Gemini Flash free-tier daily requests250 RPD20 RPDDeveloper report, Dec 6, 2025, with Google's reply on Dec 7
Free image generation modelsAvailableNone on the free tierDeveloper reports, Dec 12, 2025

Google did not publish a before-and-after table, so the numbers above are what developers saw in their own projects. Since then the free tier has kept narrowing: since September 18, 2026 only projects that used the 2.5 models before can call them, and the newest paid models (Gemini 4 Carbon, Argon and 3.1 Pro Preview) have never had a free tier.

Migration strategies that still hold:

  1. Link a billing account. Tier 1 typically takes effect instantly and moves the project to paid-tier limits, but it adds the $250 monthly cap and the $10-per-10-minutes spend limit.
  2. Move high-volume calls to Gemini 3.1 Flash-Lite or 3.5 Flash-Lite, which cost a fraction of Flash per token.
  3. Pack several items into one request where the task allows it, since RPD counts requests rather than tokens.
  4. Send non-urgent work through the Batch API or Flex at half price.
  5. Space out retries with exponential backoff. If you are already seeing 429 errors, the guide on fixing Gemini API Resource Exhausted errors walks through the causes.

What Changed in Gemini API Pricing and Models in 2026

Gemini API models, prices and billing rules changed repeatedly during 2026. The dates below come from Google's release notes and deprecations page.

DateChange
March 9, 2026Gemini 3 Pro Preview shut down; its ID points to 3.1 Pro Preview
March 12–16, 2026Project-level spend caps, revamped usage tiers and billing-account spend caps
March 23, 2026Prepay and Postpay billing plans in AI Studio; new accounts prepay at least $5
April 1, 2026Flex (50% off) and Priority inference tiers introduced
May 7, 2026Gemini 3.1 Flash-Lite generally available
June 1, 2026Gemini 2.0 Flash and 2.0 Flash-Lite shut down
July 21, 2026Gemini 3.6 Flash and 3.5 Flash-Lite generally available
September 2, 2026Gemini 3.8 Flash generally available
September 18, 20262.5 models limited to projects that used them before
October 8, 2026gemini-3.7-flash routed to 3.8 Flash; gemini-3.5-flash routed to 3.6 Flash
October 13, 2026Gemini 4 Argon generally available ($2 / $10 introductory price)
October 26, 2026New paid-tier projects store Generate Content API requests by default (store=true)
November 2, 2026gemini-2.5-pro and gemini-2.5-flash deprecated
November 10, 2026Gemini 4 Carbon generally available ($2 / $10 introductory price)
November 12, 2026Gemini 4 Hydrogen generally available ($1.50 / $7.50, free tier)
December 1, 2026Argon and Carbon rise to $4 / $20; 2.5 Pro and 2.5 Flash calls bill at 3.8 Flash and 3.5 Flash-Lite rates
January 1, 20273.8 Flash and 3.6 Flash rise to $1.50 / $7.50

Free Tier vs Paid Tiers Comparison

The free tier and the paid tiers differ in which models you can call, what Google does with your data, and which limits stop you. The price per token is either zero or the list price; there is no discount between Tier 1 and Tier 3.

Free Tier Benefits and Limitations

The free tier needs no credit card. As of November 14, 2026, the pricing page shows free input and output for six current text models: Gemini 4 Hydrogen, Gemini 3.8 Flash, Gemini 3.6 Flash, Gemini 3.5 Flash-Lite, Gemini 3.1 Flash-Lite and Gemini 3 Flash Preview. The 2.5 models also show free rows, but only for projects that used them before. Gemini 4 Carbon, Gemini 4 Argon and Gemini 3.1 Pro Preview are paid only. Using Google AI Studio itself is free in all available regions.

The costs of the free tier are not in dollars. Prompts and responses on the free tier are used to improve Google's products. The Batch API and Flex are paid features. Google does not publish free-tier RPM or RPD numbers; they appear per project in AI Studio and have changed without notice before. Per-model allowances and how to check them are covered in the Gemini API free tier guide.

Tier 1 (Pay-as-You-Go)

Linking a billing account moves the project to Tier 1; Google says this typically takes effect instantly. New accounts are put on Prepay: you buy at least $5 of credits (at most $5,000 at a time) before calls are billed. Unused credits expire after 12 months and are not refundable. When the prepaid balance reaches $0, every API key in every project on that billing account stops at once and requests fail with HTTP 402 until you add credits; there can be roughly 10 minutes of overage before usage stops. Tier 1 projects get higher rate limits than the free tier, access to all paid models, Batch, Flex and context caching, and their prompts are not used to improve Google's products. The limits are a $250 monthly cap and the $10 per-10-minute spend limit. The $300 Google Cloud welcome credit cannot pay for the Gemini API if the account was opened after March 2, 2026.

Tier 2 (Higher Limits)

Tier 2 is automatic: Google upgrades the billing account once it has paid $100 in total and 3 days have passed since its first successful payment. Spending on other Google Cloud services on the same billing account counts toward the threshold. The monthly cap rises to $2,000 and the 10-minute spend limit to $200. Per-token prices stay the same; the rate limits shown in AI Studio go up.

Tier 3

Tier 3 follows after $1,000 paid and 30 days since the first payment. The monthly cap is $20,000 to $100,000 or more, and you can ask Google to raise any tier's cap with its request form. Google notes that a qualifying upgrade can occasionally be refused after review.

Enterprise Tier

Enterprise use runs through Gemini Enterprise Agent Platform (Vertex AI), which Google describes as offering dedicated support, provisioned throughput and volume-based discounts "based on usage." Prices there can differ from the Gemini Developer API list, and Google publishes no fixed discount level. Search grounding on Vertex AI has a default quota of 1,000,000 queries a day instead of 1,500.

For what the free tier allows per model, including the Pro question, see the Gemini API free tier guide.

Monthly bill for an example workload of 200M input and 100M output tokens by Gemini model, against the $250 Tier 1 cap and the $2,000 Tier 2 cap: $200 on 3.1 Flash-Lite, $310 on 3.5 Flash-Lite, $525 then $1,050 on 3.8 Flash, $1,050 on Gemini 4 Hydrogen, $1,400 then $2,800 on Gemini 4 Argon or Carbon, $1,600 on 3.1 Pro Preview

Cost Calculator and Real-World Examples

A Gemini API bill is input tokens × input price + output tokens × output price, per million tokens, plus tool fees and cache storage. Compare the result with your tier's monthly cap: that tells you whether the bill will be paid or the project will pause.

Monthly cost = (input tokens ÷ 1,000,000 × input price) + (output tokens ÷ 1,000,000 × output price) + grounding and storage charges

Token Counting Fundamentals

Google's token guide says a Gemini token is about 4 characters, and 100 tokens are about 60–80 English words, so 1,000 words is roughly 1,300–1,700 tokens. An image of up to 384 pixels on both sides counts as 258 tokens; larger images are cut into 768 × 768 tiles of 258 tokens each. Video costs 263 tokens per second and audio 32 tokens per second. Thinking tokens are billed as output, so a reasoning model can bill several times more output than the visible answer.

Light Usage Scenario: Personal Projects

A personal assistant answering 50–100 questions a day at about 1,000 tokens each uses about 2 million input and 1 million output tokens a month. On Gemini 3.1 Flash-Lite:

Input:  2M tokens × $0.25/M = $0.50
Output: 1M tokens × $1.50/M = $1.50
Total:  $2.00/month

The same usage cost $0.60 on 2.5 Flash-Lite in January, but new projects can no longer call that model. 3.1 Flash-Lite has a free tier, so this workload may cost nothing if your project's daily request limit in AI Studio covers about 100 requests a day; if it does not, Tier 1 handles it far below the $250 cap.

Medium Usage Scenario: Startup MVP

A customer support bot handling 500–1,000 conversations a day at about 2,000 tokens each processes roughly 30 million input and 15 million output tokens a month. On Gemini 3.5 Flash-Lite:

Input:  30M tokens × $0.30/M = $9.00
Output: 15M tokens × $2.50/M = $37.50
Total:  $46.50/month

That is exactly what 2.5 Flash cost for this workload in January, and after December 1 it stays the same for projects still calling gemini-2.5-flash, because those calls are billed at 3.5 Flash-Lite rates. A long system prompt that hits the implicit cache bills the cached part at $0.03 instead of $0.30.

Heavy Usage Scenario: Production SaaS

A production application with 5,000+ requests a day at about 3,000 tokens each processes roughly 200 million input and 100 million output tokens a month. At January 2026 prices, a mix of 30% Gemini 2.5 Pro and 70% 2.5 Flash cost:

Pro (complex queries - 30%):
  Input:  60M × $1.25/M = $75
  Output: 30M × $10.00/M = $300

Flash (standard queries - 70%):
  Input:  140M × $0.30/M = $42
  Output: 70M × $2.50/M = $175

Total: $592/month

If that code still calls the 2.5 models, the same token volumes bill differently after December 1, when 2.5 Pro calls go to 3.8 Flash and 2.5 Flash calls go to 3.5 Flash-Lite:

December 2026:  60M × $0.75 + 30M × $3.75 = $157.50  (2.5 Pro → 3.8 Flash)
                140M × $0.30 + 70M × $2.50 = $217.00  (2.5 Flash → 3.5 Flash-Lite)
                Total: $374.50/month

From January 1, 2027: 60M × $1.50 + 30M × $7.50 = $315.00
                      + $217.00
                      Total: $532.00/month

These figures assume the new models use the same number of tokens per request, which a different model may not do. Either way the bill is above Tier 1's $250 cap, so the billing account needs Tier 2 ($100 paid and 3 days) before the month's usage passes $250; otherwise all projects on it pause for the rest of the month.

Monthly cost by model and the tier each bill needs (Standard prices, same three workloads):

ModelLight: 2M in / 1M outMedium: 30M in / 15M outHeavy: 200M in / 100M out
Gemini 3.1 Flash-Lite$2.00$30.00$200.00 (Tier 1)
Gemini 3.5 Flash-Lite$3.10$46.50$310.00 (Tier 2)
Gemini 3.8 Flash or 3.6 Flash, until Dec 31$5.25$78.75$525.00 (Tier 2)
Gemini 3.8 Flash or 3.6 Flash from Jan 1; Gemini 4 Hydrogen$10.50$157.50$1,050.00 (Tier 2)
Gemini 4 Argon or Carbon, until Nov 30$14.00$210.00$1,400.00 (Tier 2)
Gemini 4 Argon or Carbon from Dec 1$28.00$420.00 (Tier 2)$2,800.00 (Tier 3)
Gemini 3.1 Pro Preview (prompts ≤200K)$16.00$240.00 (Tier 1, near the cap)$1,600.00 (Tier 2)
Gemini 2.5 Flash-Lite (past users only)$0.60$9.00$60.00

Cells without a tier fit in Tier 1's $250 monthly cap. Batch or Flex halves any row for work that can wait. A heavy workload on Argon after December 1 needs Tier 3 ($1,000 paid and 30 days), because $2,800 is above Tier 2's $2,000 cap.

Python Cost Estimation Code

The function below uses the Standard prices listed above (as of November 14, 2026) and reports the lowest tier whose monthly cap covers the bill. Update the PRICING table when Google's prices change.

python
# Standard prices per 1M tokens (input, output), USD, as of November 14, 2026
PRICING = {
    "gemini-3.1-flash-lite": (0.25, 1.50),
    "gemini-3.5-flash-lite": (0.30, 2.50),
    "gemini-3.8-flash": (0.75, 3.75),         # $1.50 / $7.50 from 2027-01-01
    "gemini-4-hydrogen": (1.50, 7.50),
    "gemini-4-argon": (2.00, 10.00),          # $4.00 / $20.00 from 2026-12-01
    "gemini-3.1-pro-preview": (2.00, 12.00),  # prompts up to 200K tokens
}

# Monthly spend cap per usage tier (billing account level)
TIER_CAPS = [("Tier 1", 250), ("Tier 2", 2_000), ("Tier 3", 20_000)]


def estimate_monthly_cost(
    daily_requests: int,
    avg_input_tokens: int,
    avg_output_tokens: int,
    model: str = "gemini-3.8-flash",
    batch: bool = False,
) -> dict:
    input_price, output_price = PRICING[model]
    if batch:  # Batch API and Flex are billed at 50%
        input_price, output_price = input_price / 2, output_price / 2

    monthly_input = daily_requests * avg_input_tokens * 30
    monthly_output = daily_requests * avg_output_tokens * 30
    input_cost = monthly_input / 1_000_000 * input_price
    output_cost = monthly_output / 1_000_000 * output_price
    total = input_cost + output_cost

    tier = next((name for name, cap in TIER_CAPS if total <= cap), "Tier 3 (ask for a higher cap)")
    return {
        "model": model,
        "monthly_input_tokens": monthly_input,
        "monthly_output_tokens": monthly_output,
        "input_cost": round(input_cost, 2),
        "output_cost": round(output_cost, 2),
        "total_cost": round(total, 2),
        "lowest_tier_that_covers_it": tier,
    }


result = estimate_monthly_cost(
    daily_requests=500,
    avg_input_tokens=2000,
    avg_output_tokens=1000,
    model="gemini-3.8-flash",
)
print(result)
# 30M input, 15M output: $22.50 + $56.25 = $78.75, fits Tier 1

Choosing the Right Model for Your Use Case

Pick the cheapest model that does the job, then check the bill against your tier cap. Google's own descriptions give a starting point: 3.1 Flash-Lite and 3.5 Flash-Lite are its cost-efficient models for high-volume tasks, 3.8 Flash is its most capable Flash model for long software-engineering and agent work, Gemini 4 Hydrogen brings Gemini 4 to everyday coding and agent workflows with a free tier, Argon is the faster Gemini 4 reasoning model, and Carbon is the most capable one.

Chatbots and Conversational AI

General chatbots, FAQ answers and classification usually run on a Flash-Lite model. 3.1 Flash-Lite's $1.50 output price is 40% of 3.8 Flash's $3.75 today and a fifth of its $7.50 from January. Move a conversation up to 3.8 Flash only when the cheaper model's answers fail your own quality checks.

Document Analysis and Summarization

Long documents are mostly input tokens, so the input price and input TPM matter more than output. 3.8 Flash at $0.75 input (until December 31) keeps reading cheap: a 100,000-token document costs about $0.08 of input. On 3.1 Pro Preview, a prompt above 200,000 tokens moves the whole request to the $4.00 / $18.00 rate.

Code Generation and Review

Code agents produce long outputs and many thinking tokens, which makes output price the main cost. Gemini 4 Hydrogen ($7.50 output) and Argon ($10.00 output, $20.00 from December 1) are the Gemini 4 options; 3.8 Flash is the lower-cost one. Simple completions and formatting do not need a reasoning model.

Content Generation

Marketing copy and blog drafts work on Flash or Flash-Lite. A Gemini 4 model is worth its price for long documents where structure and consistency matter more than speed.

Real-Time Applications

Live support, games and translation need low latency. Flash-Lite models answer fastest at the lowest price. If you need guaranteed response times under load, Priority inference costs 75–100% more and has lower rate limits (0.3 times standard).

Decision Framework Table

Monthly cost at 1,000 requests a day, assuming 1,500 input and 500 output tokens per request (45 million input and 15 million output tokens a month), Standard prices as of November 14, 2026:

Use CaseRecommended ModelMonthly Cost (1000 req/day)
Simple chatbot3.1 Flash-Lite$33.75
Customer support3.5 Flash-Lite or 3.8 Flash$51.00 / $90.00 ($180.00 from Jan 1)
Document analysis3.8 Flash$90.00 ($180.00 from Jan 1)
Code generationGemini 4 Hydrogen or Argon$180.00 / $240.00 ($480.00 from Dec 1)
Real-time apps3.1 Flash-Lite$33.75
Research/reasoningGemini 4 Carbon or 3.1 Pro Preview$240.00 ($480.00 from Dec 1) / $270.00

Rows at $250 and above exceed Tier 1's monthly cap.

Cost Optimization Strategies

The largest savings come from choosing the inference option and the model, not from rewriting prompts. These are the levers Google documents, with what each one saves.

Batch Processing (50% Savings)

The Batch API halves the price of every supported model and returns results within 24 hours. It suits bulk content generation, overnight document processing and data enrichment. Batch jobs have their own limits (100 concurrent jobs, 2 GB input files, and a per-model queue of enqueued tokens), separate from your interactive RPM.

Flex Inference (50% Savings)

Flex, introduced on April 1, 2026, gives the same 50% discount for synchronous calls that can wait minutes (a 1–15 minute target) and may be dropped when capacity is short. It fits background agents and evaluations that cannot be packed into a batch file.

Context Caching (Up to 90% Savings on Cached Input)

When prompts repeat the same content (system prompts, reference documents, few-shot examples), cached input tokens are billed at 10% of the normal input price, plus storage per million tokens per hour. Implicit caching is automatic on 2.5 and newer models once a prompt reaches 4,096 tokens (2,048 on 2.5). Example: a 10,000-token system prompt sent with each of 100,000 monthly requests on 3.8 Flash costs $750 at full input price ($0.75 × 1,000 million tokens) and $75 when every request hits the cache, before storage charges. For a full walkthrough, see the Gemini API context caching guide.

Prompt Optimization

Shorter prompts cost less and use less input TPM. Remove repeated instructions, ask for structured output (a JSON schema keeps answers short), use templates instead of rewriting the whole prompt, and send only the conversation history the next answer needs. Set a thinking level that matches the task; thinking tokens are billed as output.

Model Selection Optimization

Not every request needs your most capable model. Start with a Flash-Lite model, escalate to 3.8 Flash if a quality check fails, and keep Gemini 4 for requests that need it. A routing step that sends 80% of traffic to 3.1 Flash-Lite instead of 3.8 Flash cuts that share's output cost by 60% today and 80% from January 1.

Token Efficiency Techniques

Summarize long inputs before the main call, drop old conversation turns, use structured data instead of prose where possible, and keep frequent lookups (product catalogs, user profiles) in your own database instead of every prompt.

Watch the Spend Caps

Set a project-level spend cap in AI Studio if several projects share one billing account, and turn on Prepay auto-reload with a monthly auto-charge limit so a credit balance of $0 does not stop every key at once. Spend caps can overrun by about 10 minutes because billing data arrives with a delay.

API Aggregation for Cost Management

For teams that use several AI providers, gateways such as laozhang.ai accept OpenAI, Anthropic Messages and Gemini-format requests with one API key and bill a single prepaid balance, so one account covers Gemini alongside OpenAI or Claude models. Its current per-model rates and billing rules are listed in its documentation at https://docs.laozhang.ai/.

FAQ and Conclusion

Is Gemini API really free?

Partly. As of November 14, 2026, six current text models have a free tier: Gemini 4 Hydrogen, 3.8 Flash, 3.6 Flash, 3.5 Flash-Lite, 3.1 Flash-Lite and 3 Flash Preview. Gemini 4 Carbon, Gemini 4 Argon and 3.1 Pro Preview are paid only. Free-tier limits are set per project and shown only in AI Studio; after the December 2025 cuts developers reported about 20 requests a day on Flash, and Google calls the free tier best effort. Free-tier prompts are used to improve Google's products. It suits testing and light personal projects, not production.

How much does 1 million tokens cost in Gemini?

It depends on the model: from $0.25 input and $1.50 output on 3.1 Flash-Lite to $2.00 input and $10.00 output on Gemini 4 Argon and Carbon, which rise to $4.00 and $20.00 on December 1, 2026. Gemini 3.8 Flash costs $0.75 and $3.75 until December 31, then $1.50 and $7.50. Batch and Flex halve these prices.

How does Gemini pricing compare to OpenAI?

At the low end they are close, and at the top they are equal for now. As of November 14, 2026, OpenAI lists GPT-6 Luna at $0.10 input and $0.50 output and GPT-6.1 Sol at $2.00 and $10.00 per million tokens (OpenAI pricing). Gemini 4 Argon matches Sol at $2 / $10 until November 30 and costs twice as much from December 1. For a new project, Gemini's cheapest general model (3.1 Flash-Lite, $0.25 / $1.50) costs more per token than Luna. GPT-4o is still listed at $2.50 / $10.00, not the $5.00 input price quoted in older comparisons. A full three-vendor comparison is in Gemini API vs OpenAI vs Claude cost guide.

What happens when I exceed my quota?

The API returns HTTP 429 RESOURCE_EXHAUSTED. RPM and TPM recover within the minute, RPD resets at midnight Pacific Time, and the spend limit is a rolling 10-minute window. Retry with exponential backoff instead of retrying at once. Reaching your tier's monthly spend cap pauses all projects on the billing account until the next month, and an empty Prepay balance returns HTTP 402 until you add credits.

Can I upgrade from free to paid tier instantly?

Yes. Linking an active billing account in AI Studio moves the project to Tier 1, typically instantly, with no approval step; new accounts buy at least $5 of prepaid credit during setup. Tier 2 and Tier 3 are applied automatically, usually within 10 minutes, once the billing account reaches $100 paid plus 3 days, or $1,000 paid plus 30 days.

Do prompts on the free tier train Google's models?

Yes. Google's terms say free-tier content is used to improve its products, and paid-tier content is not. Note that since October 26, 2026, new paid-tier projects store Generate Content API requests by default; you can turn this off per request (store=false) or in the project's logging settings. If data privacy is critical, use Tier 1 or higher and check that setting.

How do I monitor my API spending?

The AI Studio billing page shows your balance, usage and payments, and the Google Cloud console's billing reports break costs down by project and service. Set a project spend cap in AI Studio and a monthly auto-charge limit for Prepay auto-reload. Billing data arrives with about a 10-minute delay, so a fast loop can overrun a cap slightly.

What changes on December 1, 2026?

Gemini 4 Argon and Carbon go from $2 / $10 to $4 / $20 per million tokens (cached input from $0.10 to $0.20). Requests to gemini-2.5-pro start billing at Gemini 3.8 Flash rates and requests to gemini-2.5-flash at Gemini 3.5 Flash-Lite rates. On January 1, 2027, 3.8 Flash and 3.6 Flash go from $0.75 / $3.75 to $1.50 / $7.50.

Is there a spending limit on Gemini API Tier 1?

Yes. A Tier 1 billing account can spend at most $250 a month on the Gemini API and $10 in any 10-minute window. Above either limit, requests fail until the window passes, the month resets, the account reaches Tier 2 ($100 paid and 3 days), or Google approves a higher cap.

Can the $300 Google Cloud credit pay for the Gemini API?

Not for accounts opened after March 2, 2026. Google's billing guide says Gemini API usage is excluded from the $300 Google Cloud free trial from March 2026; the credit can still pay for other eligible Google Cloud services.

Key Takeaways

As of November 14, 2026: the cheapest model a new project can use is 3.1 Flash-Lite at $0.25 / $1.50 per million tokens; Gemini 3.8 Flash costs $0.75 / $3.75 until December 31; Gemini 4 Argon and Carbon cost $2 / $10 until November 30 and $4 / $20 after. Six text models have a free tier, but its limits are unpublished and have been cut before. On paid tiers, the monthly cap ($250 / $2,000 / $20,000+) and the 10-minute spend limit are often reached before any per-minute rate limit. Batch and Flex halve the bill, and context caching takes 90% off repeated input.

For current Gemini API documentation, see Google's pricing, rate limits and billing pages. For handling limits and errors, see Gemini API rate limits by tier and, for the AI Studio message, "You've reached your rate limit" in AI Studio.

Found an error, or hit a problem this article does not cover? Send the article URL and what you saw to hi@laozhang.ai. We check it and update the article.