Gemini API Pricing 2026: Every Model Rate and Tier
Current Gemini API pricing per million tokens, from 2.5 Flash-Lite at $0.10/$0.40 to Gemini 4 Argon, plus the free tier, service tiers, and the price changes landing December 1 and January 1.

Gemini API pricing is billed per million tokens, and the model, the service tier and the prompt length set your rate. On the paid Standard tier, Gemini 2.5 Flash-Lite costs $0.10 input and $0.40 output, Gemini 3.8 Flash costs $0.75 and $3.75, and Gemini 4 Argon costs $2 and $10. Batch and Flex bill at half of Standard, Priority at 1.8 times Standard.
A rate card prices one request. It says nothing about how many requests you send, and that number grows with usage. Major is the enterprise platform where agents build the software they run on, so for repeatable work such as classification, extraction and routing, an agent builds a governed app once and calls Gemini only for the cases that need judgment. The build section below shows how.
Rates were read from Google's Gemini API pricing page on November 4, 2026. Google changes this page often, so verify before you budget.
Gemini API pricing by model
All figures are USD per million tokens, text input and output, paid Standard tier.
- Gemini 4 Argon: $2 input, $10 output. No free tier. Rates double on December 1, 2026, to $4 and $20.
- Gemini 3.1 Pro Preview: $2 and $12 up to a 200k-token prompt, $4 and $18 above it. No free tier.
- Gemini 3.8 Flash and 3.6 Flash: $0.75 and $3.75. Both rise to $1.50 and $7.50 on January 1, 2027.
- Gemini 3 Flash Preview: $0.50 and $3.00 for text, image and video.
- Gemini 3.5 Flash-Lite: $0.30 and $2.50.
- Gemini 3.1 Flash-Lite: $0.25 and $1.50.
- Gemini 2.5 Pro (deprecated): $1.25 and $10 up to 200k, $2.50 and $15 above.
- Gemini 2.5 Flash (deprecated): $0.30 and $2.50.
- Gemini 2.5 Flash-Lite: $0.10 and $0.40. The cheapest text rate on the card.
Audio input is priced separately on several Flash models, so check the audio line if you transcribe or process calls. Image generation, video, music and embeddings sit on other scales that this article skips.
Version number does not predict price. Gemini 3.8 Flash costs less than Gemini 3 Flash Preview at the high end only after you compare output rates, so read the row for the exact model.
Is the Gemini API free?
Partly. The page lists free inference on Standard for Gemini 2.5 Flash, 2.5 Flash-Lite, 3.8 Flash, 3.6 Flash and the 3 Flash Preview, and in all four modes for Gemini 3.1 Flash-Lite. Gemini 3.1 Pro Preview and Gemini 4 Argon have no free tier. Free usage carries lower rate limits, and Batch and Flex are unavailable on most free-tier models.
Google's terms mark free-tier content as used to improve its products and paid-tier content as not. A prototype on synthetic data can use the free tier. Anything touching customer records should start on paid, because the data terms settle the question before price does.
Limits apply per project, not per API key, and rise as your billing tier rises. Check your current tier in Google AI Studio.
Standard vs Batch vs Flex vs Priority
Tier choice is the most under-used lever on the card. On every model that offers them, Batch and Flex bill at half of Standard and Priority bills at 1.8 times Standard. Argon is the clearest example: $2 and $10 Standard, $1 and $5 on Batch and Flex, $3.60 and $18 on Priority.
- Standard: synchronous default, for when a person or live system is waiting.
- Batch: half price, results returned asynchronously. Backfills, overnight jobs, evaluation runs.
- Flex: half price at lower priority with variable latency. Background work that tolerates a slow answer.
- Priority: 1.8 times Standard for faster, steadier service. Use it when latency protects revenue or an SLA.
Moving a non-interactive job from Standard to Batch halves the bill with identical output. Batch and Flex are marked unavailable on several models, so confirm the model before you design around the half-price rate.
What does the 200,000-token threshold do?
Pro-class models have two bands. Gemini 3.1 Pro Preview moves from $2 to $4 input and from $12 to $18 output when the prompt passes 200k tokens. Gemini 2.5 Pro moves from $1.25 to $2.50 input. The threshold applies to the whole prompt, so a long system prompt plus retrieved context can cross it even when no document looks large. Flash and Flash-Lite rows have no such band.
What does context caching cost?
Cached input is its own line at a fraction of normal input. Gemini 2.5 Flash-Lite lists cached text input at $0.01 per million on Standard, against $0.10 for fresh input. Gemini 3.8 Flash lists $0.075 against $0.75. Several models also charge a storage fee per million tokens per hour, so caching pays off when the same system prompt and reference material go out on many calls. Gemini 4 Argon supports implicit caching only.
What changes on December 1 and January 1?
Two increases and one reroute are on the page.
- December 1, 2026: Gemini 4 Argon Standard doubles to $4 input and $20 output. Batch and Flex become $2 and $10. Priority becomes $7.20 and $36.
- December 1, 2026: requests to Gemini 2.5 Pro route to Gemini 3.8 Flash, and requests to Gemini 2.5 Flash route to Gemini 3.5 Flash-Lite, each billed at the destination model's rates.
- January 1, 2027: Gemini 3.8 and 3.6 Flash move to $1.50 and $7.50 Standard, $0.75 and $3.75 Batch and Flex, $2.70 and $13.50 Priority.
If you budget twelve months on any of these models, the second half can cost twice the first. Pinned model names that silently reroute also change your output, not only your bill.
What does a real job cost?
Take a nightly job that classifies 40,000 support tickets on Gemini 3.8 Flash, at 1,500 input and 200 output tokens per ticket. That is 60M input and 8M output tokens per run.
Standard costs $45 for input and $30 for output, so $75 per run and $2,250 over 30 days. Batch halves that to $37.50 per run. After January 1, 2027, the same runs cost $150 on Standard and $75 on Batch. Nobody waits on an overnight job, so Standard was paying for responsiveness with no value. The ratios hold even when your token counts differ.
Build the ticket classification agent on Major
Batch at half price on a workload that triples is still a line that goes up. Every tier assumes the model does the work on each request. Major moves the repeatable part off the rate card, and it is the first place to look when Gemini spend scales with volume.
- Connectors. Zendesk supplies tickets and Slack carries escalations. Credentials are scoped per connector, so the agent reads tickets and posts to one channel and nothing else.
- The app. The agent works out the classification rules once, then builds an app that applies them in deterministic code. The app has a managed database for categories, rule versions and run history, and it writes its own audit log.
- The agent's judgment. Tickets the rules cannot place go to a Gemini call, typically a Flash-Lite model on Batch. The model makes only the decisions that need judgment.
- State. Account history and reference context live in the app's database, so they are not re-sent as input tokens on every run.
- Trigger. A scheduled workflow runs the app nightly. Ambiguous or high-risk tickets route to a person in Slack for approval.
- Governance. Role-based access controls who edits rules, and every classification is attributed and logged.
The model still reasons. It writes the app, handles exceptions and flags drift. Cost is front-loaded and then flat, because building the app spends reasoning up front and later runs stop scaling token spend with ticket volume. Reason once, run forever. Start building on Major and point it at your own queue.
For the Zendesk side of this pattern, see how to build a triage-and-draft agent you control.
How does Gemini pricing compare with other model APIs?
The same structure, per-token billing with tier and caching levers, shows up across providers, and the lowest sticker price rarely gives the lowest bill. Our cheapest LLM API breakdown walks through why, and Anthropic vs OpenAI for builders covers the two other APIs teams most often price against Gemini. If you are choosing which work should hit a model at all, AI agent use cases shows which workflows are repeatable enough to become apps.
Start building on Major
Build a ticket classification agent on Major that applies its rules in a governed app and calls Gemini only for the tickets that need judgment, so token spend stops tracking ticket volume.
Related articles
Frequently asked questions
- How much does the Gemini API cost?
- Gemini API pricing is per million tokens and depends on model, service tier and prompt length. On the paid Standard tier, Gemini 2.5 Flash-Lite costs $0.10 input and $0.40 output, Gemini 3.8 Flash costs $0.75 and $3.75, and Gemini 4 Argon costs $2 and $10. Batch and Flex cost half of Standard. Checked November 4, 2026.
- Is the Gemini API free to use?
- Partly. Google lists free Standard inference for Gemini 2.5 Flash, 2.5 Flash-Lite, 3.8 Flash and 3.6 Flash, with lower rate limits. Gemini 3.1 Pro Preview and Gemini 4 Argon have no free tier. Free-tier content may be used to improve Google's products, so avoid customer data there.
- What is Gemini 2.5 Flash-Lite pricing?
- Gemini 2.5 Flash-Lite costs $0.10 per million input tokens and $0.40 per million output tokens on the paid Standard tier for text, image and video. Batch and Flex are $0.05 and $0.20, Priority is $0.18 and $0.72, and audio input is priced separately at $0.30 on Standard.
- Is Gemini 2.5 Flash free of charge?
- Google lists Gemini 2.5 Flash inference as free on Standard and Priority modes, with lower rate limits, and Batch and Flex unavailable on the free tier. Paid Standard use costs $0.30 input and $2.50 output per million tokens. The model is marked deprecated, and requests route to Gemini 3.5 Flash-Lite on December 1, 2026.
- What is the difference between Standard, Batch, Flex and Priority?
- They price the same model at different service levels. Standard is the synchronous default. Batch and Flex bill at half of Standard, with asynchronous or lower-priority delivery. Priority bills at 1.8 times Standard for faster, steadier service. Not every model offers every mode.
- When do Gemini API prices go up?
- Google lists two increases. Gemini 4 Argon Standard doubles to $4 input and $20 output on December 1, 2026. Gemini 3.8 and 3.6 Flash rise to $1.50 input and $7.50 output on January 1, 2027. Re-check the pricing page before you lock a budget.